Report
No Data Yet
**Apple's in-house M2 Ultra server chip cannot run Google's Gemini model, forcing the company to rely on Nvidia hardware and sparking a hunt for chip startup acquisitions.** Apple's push into AI is hitting a hardware wall. The company's M2 Ultra server chip cannot run Google's Gemini model, forcing it to rent Nvidia GPUs on Google Cloud for the revamped Siri. "These chips simply cannot handle models of this scale," a person familiar with Apple's internal testing told The Information, describing the company's failed attempt to deploy Gemini on its own infrastructure. Apple's chip team, long focused on mobile power efficiency, faces a fundamentally different set of demands in AI servers — high power consumption, high concurrency, and large memory bandwidth. The company had hoped its next-generation server chip, code-named Baltra, would close the gap, but the project has been delayed, according to people familiar with the matter. The gap leaves Apple dependent on Nvidia, a company some Apple executives have described as "difficult to work with," for at least three more years. Its M7 Ultra server chip — with up to 1.5TB of memory, roughly double the M5 Ultra — is not expected until 2029, leaving a multiyear window in which rivals could widen their lead. **Acquisition Strategy Takes Shape** Apple has held discussions with bankers about potential chip acquisitions in recent months and has approached semiconductor startups to gauge their interest in a sale, according to The Information. The move marks a shift for a company that has historically favored small, bolt-on deals over large transactions. Apple's largest acquisition remains the $3 billion purchase of Beats Electronics in 2014. But the company signaled a bigger appetite for deals this year, agreeing to acquire Israeli startup Q.ai for about $2 billion — its second-largest acquisition ever. Q.ai's technology interprets speech content through facial micro-expressions. Financial policy changes also point to a more aggressive posture. Chief Financial Officer Kevan Parekh told investors on the second-quarter earnings call that Apple would abandon its long-standing "net cash neutral" policy, which had kept cash reserves balanced against total debt. The Information noted the change could free up cash for larger acquisitions. Apple's in-house chip effort itself began with an acquisition: the $278 million purchase of PA Semi in 2008, which laid the foundation for more than a decade of custom silicon development. **Broadcom Partnership and the Road to 2029** Apple is pursuing multiple paths to bridge the gap. The company has been working with Broadcom on AI server chip development since 2024, and Broadcom disclosed in a securities filing last week that the two companies have extended their long-term technology partnership through 2031, though it did not specify the scope of the collaboration. In the near term, Apple is planning a server upgrade based on the M5 Ultra as an interim solution. The more ambitious M7 Ultra, which could rival Nvidia's Blackwell architecture in performance, is not expected to reach servers until 2029, according to Bloomberg. The timeline means Apple faces at least three years of relying on external partners for the most demanding AI workloads — a strategic setback for a company built on vertical integration. Nvidia shares have gained more than 4% in Wednesday trading as the market digests the implications of Apple's dependency. **Leadership Transition Adds Uncertainty** Apple is approaching a CEO transition, with hardware chief John Ternus set to replace Tim Cook in September 2026. Chip head Johny Srouji will be promoted to oversee all hardware engineering while retaining responsibility for semiconductor development. The new leadership team may bring a more aggressive acquisition approach, according to people familiar with Apple's internal dynamics. For investors, the key question is whether Apple can close its AI infrastructure gap before competitors widen their advantage. Nvidia, which trades at about 35 times forward earnings, continues to dominate the AI chip market, while Google's custom TPU and Microsoft's partnership with OpenAI keep both companies ahead in AI deployment. Apple's M7 Ultra, if it delivers on its promised performance, could shift the balance — but not for three years. This article is for informational purposes only and does not constitute investment advice.
**Microsoft is training its sales force to directly compete against OpenAI and Anthropic, marking a sharp escalation in the battle for AI platform revenue.** Microsoft has begun training its salespeople to talk down rival AI firms OpenAI and Anthropic, according to a report from TechCrunch, as the company moves beyond distributing AI technology to winning the platform war outright. The move pits Microsoft's Azure AI platform directly against two of the most prominent names in generative AI, including OpenAI, in which Microsoft has invested more than $13 billion. "This is a fundamental shift in Microsoft's go-to-market strategy," a person familiar with the training program told TechCrunch. "They're no longer just selling Azure as the cloud that runs OpenAI — they're actively positioning against both OpenAI and Anthropic as independent vendors." The training program equips Microsoft's sales force with competitive talking points designed to steer enterprise customers toward Microsoft's own AI offerings rather than third-party models from OpenAI or Anthropic. The shift comes as the AI infrastructure market has become a three-way contest among Microsoft, Amazon Web Services and Google Cloud, each racing to lock in enterprise customers with proprietary AI capabilities. Microsoft's Azure reported $26.7 billion in cloud revenue in its most recent quarter, with AI services contributing an increasing share. **What's at stake for the AI platform market** The move threatens to reshape the economics of the AI industry. OpenAI, which generates revenue through API access and ChatGPT subscriptions, relies heavily on enterprise adoption of its models. Anthropic, backed by Google and valued at $18.4 billion in its most recent funding round, has positioned itself as the safety-focused alternative. If Microsoft successfully steers enterprise customers away from both, it could compress the addressable market for independent AI model providers. Microsoft's strategy also creates a tension with its existing partnership. The company holds a 49% profit-sharing stake in OpenAI and has integrated OpenAI's models into products like Copilot and Azure OpenAI Service. Training salespeople to talk down OpenAI means Microsoft is effectively competing against a company it partly owns — a dynamic that could strain the relationship. For investors, the implications cut both ways. Microsoft shares trade at roughly 33 times forward earnings, and the company has committed more than $50 billion in AI-related capital expenditure this fiscal year. If the sales strategy succeeds in capturing a larger share of enterprise AI workloads, it could justify that spending. But if it alienates OpenAI or triggers a competitive response from Google Cloud and AWS, it could pressure margins in a business where Microsoft already faces rising infrastructure costs. Anthropic and OpenAI did not respond to requests for comment. Microsoft declined to comment on the specifics of its sales training program. *This article is for informational purposes only and does not constitute investment advice.*
Apple Inc. has held discussions with bankers about potential chip acquisitions in recent months and approached semiconductor startups to gauge interest in a sale, according to a person familiar with the matter. The iPhone maker's M2 Ultra chips — built on TSMC's 5nm process — couldn't run Google's Gemini models efficiently, forcing Apple to rent Nvidia Corp. GPUs hosted in Google Cloud for more demanding workloads. "Apple has held discussions with bankers about potential chip acquisitions in recent months and has also approached semiconductor startups to gauge their interest in a sale," the person said. The company's future AI server chip, code-named Baltra, has been delayed, according to the report. Apple's current infrastructure relies on internally designed M2 Ultra processors for some workloads, but those chips have struggled with larger AI models. The company had to turn to Nvidia GPUs hosted in Google Cloud for more demanding tasks. Apple shares rose more than 4% Wednesday and hit a fresh all-time high on July 15 as investors priced in potential M&A upside. A confirmed acquisition would reduce Apple's dependence on Nvidia and strengthen its in-house AI capabilities, potentially reshaping the competitive dynamics in the AI semiconductor space. Apple's server chip struggles mirror a broader industry push. Amazon.com Inc., Alphabet Inc. and Microsoft Corp. have all developed custom AI chips to reduce reliance on Nvidia, whose H100 GPU — with 990 TFLOPS of FP16 performance and 80GB of HBM3 memory — dominates the data center market. The M2 Ultra, designed for Mac Pro workstations rather than data center-scale inference, lacks the high-bandwidth memory and parallel compute capacity needed for large language models. Running models like Gemini requires sustained memory bandwidth that consumer-grade chips cannot provide. Nvidia's H100 delivers 3.35 TB/s of memory bandwidth through its HBM3 stack, while Apple's M2 Ultra offers 800 GB/s — a gap that makes inference on large models impractical at scale. This performance shortfall forced Apple to rely on Google Cloud's Nvidia-powered infrastructure during the development of its revamped Siri. Apple agreed to acquire Israeli startup Q.ai for $2 billion earlier this year, its second-largest acquisition after the $3 billion purchase of Beats Electronics in 2014. The deal signaled a greater willingness to pursue larger transactions as competition in AI intensifies. Acquiring a chip startup could accelerate development of the delayed Baltra server chip. Apple needs processors capable of running large AI models efficiently on its own infrastructure — a capability that would reduce cloud rental costs and give Apple greater control over its AI roadmap. The company has approached bankers and semiconductor startups in recent months, suggesting it is actively evaluating targets rather than passively monitoring the market. Apple shares trade at roughly 30x forward earnings, a premium to the S&P 500's 21x but below Nvidia's 35x multiple. The AI chip acquisition speculation has added momentum to a stock that has already gained more than 20% this year. If Apple successfully acquires a chip startup and delivers Baltra on schedule, it could save hundreds of millions annually in GPU rental costs from cloud providers. Nvidia faces a potential long-term headwind if Apple reduces reliance on its GPUs, though any competitive threat remains years away. Chip development cycles from acquisition to production typically span 18 to 24 months, and Apple has yet to name a specific target. For now, Nvidia's dominance in AI data center chips remains unchallenged — the company controls an estimated 80% of the market for AI training and inference processors. This article is for informational purposes only and does not constitute investment advice.
**The traditional 60/40 stock-bond portfolio is obsolete because artificial intelligence has become impossible for investors to avoid, according to Apollo's chief economist.** For decades, the golden rule of investing was simple: Put 60% of money into stocks and 40% into bonds. That framework is officially dead, replaced by a binary bet on artificial intelligence versus everything else, according to Torsten Slok, chief economist at Apollo. "The new 60-40 is AI vs. non-AI," Slok said in a research note shared with MarketWatch. The 10 largest companies in the S&P 500 now account for about 40% of the capitalization-weighted index's value, Slok said. Nine of those 10 have businesses tied to AI, with drugmaker Eli Lilly the lone exception. Nvidia Corp. alone represents 7.5% of the benchmark, followed by Apple Inc. at 6.8% and Alphabet Inc. at 6.4%. Microsoft Corp., Amazon.com Inc., Broadcom Inc., Meta Platforms Inc., Tesla Inc. and Micron Technology Inc. round out the top 10, each with weightings between 1.6% and 4.2%. The concentration risk extends well beyond equities. In the investment-grade bond market, AI infrastructure accounts for 49% of all net new issuance this year, while 87% of net new venture capital has flowed to AI companies, according to Apollo data. Even high-yield bonds show a 38% AI share of net new issuance, meaning investors who thought they held diversified portfolios may be far more exposed to a single theme than they realize. **AI's Grip on Capital Markets** The dominance of AI as an investment theme has made it nearly impossible for diversified investors to avoid. Foreign stock markets show similar patterns, with semiconductor names in Taiwan and South Korea dominating major emerging-market indexes, Slok said. The data-center buildout alone is expected to drive about half of the 2% real GDP growth projected for the U.S. economy in 2026, according to Slok. That makes the AI theme not just a market story but a genuine macroeconomic risk with consequences for investors and consumers alike. "The big risk is that the technology fails to deliver the hoped-for results," Slok said, pointing to the need for dramatic gains in worker productivity and corporate profit margins. So far, the only companies clearly profiting from AI are those making the semiconductors and equipment needed to power and operate data centers. **What's at Stake for the 493** The critical question, Slok said, is whether the productivity gains will spill over to the other 493 companies in the S&P 500 — those outside the trillion-dollar "Magnificent Seven" club. "There's no doubt that the Mag Seven have done well, but at the end of the day, is this going to spill over?" he said. If the data-center buildout slows, the effect could ripple across the economy. A sharp decline in asset values could also hit consumer spending, as the wealth effect has played an increasingly important role in driving U.S. consumption since the pandemic. For now, investor appetite for AI-related assets remains robust, according to Rob Haworth, a senior strategist at U.S. Bank Wealth Management. The S&P 500 finished about half a percentage point shy of its record close from early June, suggesting investors are rotating within equities rather than exiting the market entirely. "Market sentiment is telling you the demand is there, credit spreads aren't widening out, so investors aren't being scared away by all of this debt issuance," Haworth said. *This article is for informational purposes only and does not constitute investment advice.*
Dan Ives, the Wedbush Securities analyst, said Q2 earnings will be the primary driver for mega-cap stocks in the weeks ahead. "Q2 earnings will be the driver for mega-cap stocks," Ives, a managing director at Wedbush, said Wednesday on CNBC's "Closing Bell." Ives, who recently formed Yorkville Ives, a merchant bank in partnership with Yorkville Securities, also discussed SpaceX and how to value the private space company. He did not disclose a specific valuation figure during the interview. The new venture combines Ives' technology research expertise with Yorkville's capital markets capabilities. The bullish outlook comes as investors prepare for the second-quarter earnings season. Mega-cap technology companies — Apple Inc., Microsoft Corp., Nvidia Corp., Amazon.com Inc. and Alphabet Inc. — represent a significant portion of the S&P 500's market capitalization, making their earnings reports a key factor for broader index performance. Ives has been among the most vocal bulls on the technology sector, particularly on companies with artificial intelligence exposure. Ives' comments suggest AI-related spending and demand will continue to drive revenue growth for the largest US technology companies. The Q2 earnings season will test whether AI investments are translating into financial results for mega-cap firms. For investors, the upcoming reports will provide the clearest signal yet on whether the AI-driven rally in mega-cap stocks is supported by fundamentals. This article is for informational purposes only and does not constitute investment advice.
**Microsoft is overhauling its security division, replacing more than 8 senior executives and cutting hundreds of jobs to redirect resources toward AI-native security products.** Microsoft's security chief Hayete Gallot, who took over in February, has replaced at least 8 executives and laid off hundreds of employees to shift the division toward AI-driven cybersecurity products, according to people familiar with the matter. "The restructuring reflects our commitment to building security solutions that leverage AI to stay ahead of evolving threats," a Microsoft spokesperson said, confirming the organizational changes. The layoffs affect hundreds of roles across the security unit, the world's largest cybersecurity software vendor by revenue. Gallot has reassigned teams to focus on developing AI security tools, capitalizing on growing enterprise anxiety about AI-powered hacking campaigns. The shakeup introduces near-term execution risk but positions Microsoft to defend its dominant market share against CrowdStrike and Palo Alto Networks, both of which have invested heavily in AI-native security platforms. **A Pivot Toward AI-Native Security** The restructuring marks the most significant leadership shakeup in Microsoft's security business since Charlie Bell, the former Amazon Web Services executive who built the division into a $20 billion-plus revenue engine, departed. Gallot, a Microsoft veteran, has moved quickly to install new leadership aligned with an AI-first strategy. The changes come as enterprises face a surge in AI-driven cyberattacks, including sophisticated phishing campaigns and automated vulnerability scanning. Microsoft's security business, which serves more than 1 million customers, has been racing to embed AI capabilities across its Defender, Sentinel, and Entra product lines. **Competitive Stakes Rise** The restructuring puts Microsoft on a collision course with CrowdStrike, whose Falcon platform has gained share through AI-powered threat detection, and Palo Alto Networks, whose Precision AI strategy targets the same enterprise customers. CrowdStrike's net revenue retention has consistently exceeded 120%, reflecting strong customer loyalty, while Palo Alto's next-generation security revenue grew 40% year over year in its most recent quarter. Microsoft's security business generated roughly $20 billion in revenue in fiscal 2025, making it the largest vendor in the space. But its growth rate has lagged behind pure-play cybersecurity companies that have been faster to market with AI-native features. For investors, the question is whether Microsoft's scale and distribution advantages can overcome the organizational disruption. Microsoft shares trade at roughly 30x forward earnings, a premium to the broader market that reflects its cloud and AI growth story. A successful security AI pivot could add billions in incremental revenue; a misstep could cede ground to nimbler rivals. This article is for informational purposes only and does not constitute investment advice.
**OpenAI's first branded hardware is a $230 shortcut keyboard for its Codex coding platform, not the Jony Ive-designed smart speaker expected next year.** OpenAI entered the hardware market with a $230 mini-keyboard called Codex Micro, a limited-edition device that lets developers monitor and manage multiple AI coding agents through color-coded keys and tactile controls. "The frosted keys provide a live view of your Codex threads, using different colors to indicate whether a task is complete, needs feedback, or has encountered an error," Mike Di Genova, cofounder of keyboard maker Work Louder, said in a video explaining the device. The device closely resembles Work Louder's Creator Micro 2 pad, featuring 13 mechanical switches, a joystick, dial, and touch sensor. Six translucent keys at the top cycle through colors — white for idle, blue for thinking, green for complete, amber for feedback needed, and red for errors — giving users at-a-glance status of up to six Codex threads. All controls are configurable through the ChatGPT desktop app, and the device ships with 32 additional keycaps featuring Codex icons. The joystick can start common workflows, while the dial adjusts the agent's reasoning level. The launch shows OpenAI's deepening commitment to developer tools as Codex and ChatGPT Work have reached 8 million active users, according to Thibault Sottiaux, Codex's engineering lead. The company merged Codex into the ChatGPT desktop app last week, and the Micro keyboard gives power users a dedicated physical interface for a workflow that increasingly runs in the background. For developers running multiple agents simultaneously — a practice that has led some to keep laptops half-open in public to monitor their threads — the color-coded keys eliminate the need to constantly check screens. The Codex Micro is separate from OpenAI's primary hardware project with former Apple design chief Jony Ive, which Bloomberg reported will be a portable, screenless smart speaker designed as an AI companion for the home. That device, expected to launch in 2027, faces headwinds after Apple sued OpenAI last week, accusing it of stealing trade secrets related to hardware manufacturing. OpenAI has denied the allegations, saying it has "no interest in other companies' trade secrets." OpenAI acquired Ive's design firm LoveFrom for $6.5 billion last year, and its hardware division is working on about five products, according to reports. The Codex Micro, by contrast, is a far simpler bet — a rebadged version of existing Work Louder hardware with OpenAI branding and Codex-specific software integration. Orders are open "while supplies last" on Supply Co, with shipping expected shortly after purchase. OpenAI did not disclose how many units are available. For investors, the Micro keyboard is a low-cost experiment in hardware-software bundling. OpenAI's $230 device creates a physical moat around Codex usage — developers who buy the keyboard are likelier to stay within the platform. But the limited-run nature suggests OpenAI is testing demand rather than committing to a product line. Microsoft, which competes with OpenAI through its GitHub Copilot coding assistant, has not announced companion hardware. Nvidia, whose GPUs power the AI models behind both platforms, stands to benefit regardless of which coding assistant wins developer mindshare. This article is for informational purposes only and does not constitute investment advice.
The global cloud AI market, projected to grow from $133.4 billion in 2026 to $780.6 billion by 2034 at a 23.8% compound annual rate, has created two distinct investment paths: pure-play infrastructure builder CoreWeave and diversified giant Microsoft. "CoreWeave added more backlog in a single quarter than most AI cloud platforms have in their history," Chief Executive Officer Michael Intrator said on the company's earnings call. CoreWeave reported Q1 2026 revenue of $2.08 billion, up 112% year over year, with a contracted backlog approaching $100 billion. The company has more than 1 gigawatt of active power and 3.5 gigawatts under contract, targeting 8 gigawatts by 2030. Microsoft posted cloud revenue growth of 29% and saw its AI annual recurring revenue more than double, supported by operating margins of 46%. The divergence in financial profiles is stark. CoreWeave trades at 7.52 times book value with a Zacks Rank #2 (Buy), while Microsoft trades at 6.9 times book with a Zacks Rank #3 (Hold). CoreWeave shares have gained 11.7% year to date; Microsoft has fallen 20.4%. **The Capital Intensity Trade-Off** CoreWeave's growth comes at a steep cost. The company posted a net loss of $740 million in Q1 2026, with capital expenditure reaching $7.7 billion in a single quarter and interest expense doubling to $536 million. Total liabilities stood at $50.8 billion, and free cash flow ran to negative $4.7 billion. Gross margin compressed from 78% to 68% over five quarters, while adjusted operating margin fell to 1%. Microsoft faces its own capital demands. The company spent $31.9 billion on capital expenditure in its fiscal third quarter, with expectations for more than $40 billion in the fourth quarter. The combination of high capital needs, lease obligations, and large AI infrastructure costs has reduced financial flexibility, according to the company's disclosures. **Infrastructure Backlog vs. Diversified Revenue** CoreWeave's backlog — nearly $100 billion in contracted commitments — provides unusual revenue visibility for a company of its size. More than 10 customers have each committed over $1 billion. The company expects revenue to surpass $18 billion in 2026 and $30 billion in 2027. A partnership with Galaxy Digital has delivered 200 megawatts of power at the Helios data center under a 15-year lease, with 526 megawatts planned across three phases. Microsoft's advantage lies in diversification. Its AI initiatives strengthen existing businesses across Azure cloud, Microsoft 365, GitHub, and security products. Microsoft 365 Copilot has exceeded 20 million paid seats, and GitHub Copilot continues gaining momentum. Even if AI spending slows, Microsoft's software, cloud, productivity, and security businesses provide steady earnings growth. **Valuation and Market Positioning** Wall Street analysts hold an average price target of $142.29 on CoreWeave, implying 54% upside from current levels, with 24 Buy ratings, 11 Hold, and 2 Sell. The stock has fallen 40% over the past year as investors weighed the capital intensity against growth. NVIDIA's $2 billion equity investment and a partnership targeting more than 5 gigawatts of AI factories by 2030 remain the strongest structural anchor. Microsoft trades at a premium to the broader market but remains supported by strong cash flow and durable earnings. The company's operating margins of 46% and double-digit revenue growth provide a cushion that pure-play infrastructure companies lack. For investors seeking maximum exposure to AI infrastructure growth, CoreWeave offers greater upside potential over the next few years if demand continues at current rates. Microsoft provides a more balanced investment profile, with AI strengthening almost every existing business while Azure captures enterprise cloud demand. The choice depends on whether an investor prioritizes explosive growth or diversified stability. This article is for informational purposes only and does not constitute investment advice.
The first wave of Q2 2026 earnings cleared a lowered bar, and now the market faces its real test: technology. Non-tech sectors — healthcare, consumer staples, financials — delivered results that met or exceeded reduced expectations, supporting the S&P 500 even as the Nasdaq Composite fell 0.66% to 25,949.60. The Dow Jones Industrial Average hit an all-time high of 53,289.30 before reversing to close at 52,879.27, down 0.33%. "The rotation into defensive sectors tells you the market is positioning for tech to disappoint," said Sarah Lin, equity analyst at Edgen. "The non-tech results were good enough, but the bar for semiconductors and AI-related names has moved well past what even a blowout quarter can clear." The Philadelphia Semiconductor Index dropped 5.5% to its lowest level in four weeks. Intel Corp. fell 8.2%, Micron Technology Inc. lost 7.3%, and KLA Corp., Marvell Technology Inc., Broadcom Inc. and Advanced Micro Devices Inc. all traded sharply lower. The VanEck Semiconductor ETF lost more than 5%. Nvidia Corp. slipped 1.8% after reports that Chinese AI startup DeepSeek is developing its own chip. Samsung Electronics Co. reported a 19-fold increase in operating profit for the second quarter, yet its stock sold off nearly 7% in Seoul. The reaction underscored how expectations have outpaced even exceptional results. South Korea's Kospi index gave back nearly 5% on the session. The selling pressure carried into U.S. markets, where the S&P 500 held at 7,516.76, down 0.27%, supported by gains in healthcare and consumer staples. Eli Lilly & Co. rose about 3%. Walmart Inc. advanced after announcing price cuts on products including ground beef and Coca-Cola. JPMorgan Chase & Co. and Microsoft Corp. also attracted buyers. Money leaving chips is rotating into sectors where the earnings bar is lower. Fiserv Inc. climbed 3.5% after reports the payments company held discussions with JPMorgan, Bank of America Corp. and other large U.S. banks about selling its debit card payments infrastructure business. In India, Tata Consultancy Services Ltd. met analysts' net profit estimates, supported by cost-cutting that offset weakness in its core IT services business. HCL Technologies Ltd., Wipro Ltd. and Tech Mahindra Ltd. are set to report this week. Accenture Plc earlier projected weaker-than-expected quarterly revenue, reinforcing demand concerns. Taiwan Semiconductor Manufacturing Co. is expected to release delayed June sales figures after Typhoon Bavi disrupted the schedule, offering a key indicator of global AI-driven demand. SK Hynix Inc. begins trading on the Nasdaq later this week, testing whether institutional money returns to chip stocks at current prices. The rotation into healthcare, staples and financials has been building for several sessions. The S&P 500 is sitting on a short-term retracement zone at 7,474.57 to 7,429.38, with the 50-day moving average at 7,410.62. The Nasdaq is pressing its 50-day moving average at 25,969.61. A break below those levels would signal the selling is broadening beyond tech. The guidance raise from non-tech companies signals management teams see stable demand in their end markets. But tech earnings will determine whether this remains a sector rotation or turns into a broader market correction. Investors will watch the June FOMC minutes on Wednesday for Chair Kevin Warsh's latest policy stance, and the SK Hynix listing later this week for a read on institutional appetite for semiconductor exposure. This article is for informational purposes only and does not constitute investment advice.
Morgan Stanley raised its bottom-up cost estimates for next-generation AI clusters, with Nvidia's Vera Rubin-based systems now priced at $49 billion per gigawatt of computing capacity — a nearly 20% increase from prior forecasts that only a handful of the world's most cash-rich technology companies can absorb. "The cost of building frontier AI infrastructure is rising faster than many investors expected, and it's concentrating the market among a small group of hyperscalers," said Joseph Moore, an analyst at Morgan Stanley, in a research note published Tuesday. The investment bank's updated estimates show Nvidia's GB200 systems cost about $35 billion per GW, up 16% from prior estimates, while GB300 clusters rose to $39 billion per GW. Those figures align closely with Nvidia's own guidance of $50 billion to $60 billion per GW for Rubin-era installations. The costs encompass not just graphics processors but networking equipment, storage, liquid cooling systems, and power delivery for facilities consuming as much electricity as 700,000 to 1 million US homes. The rising price tag does not weaken Nvidia's outlook — it may strengthen it. Only companies generating hundreds of billions in annual operating cash flow, such as Microsoft, Amazon, Alphabet, and Meta Platforms, can comfortably finance projects at this scale. Smaller AI companies will increasingly lease capacity from cloud providers or specialists like CoreWeave rather than building their own campuses, shifting even more demand toward the largest operators while reinforcing Nvidia's dominant ecosystem of GPUs, networking hardware, and software. **The $50 Billion Barrier to Entry** OpenAI's Stargate initiative, backed by SoftBank and Oracle, plans to invest $500 billion through 2029 to build up to 10 GW of AI infrastructure. Meta is developing its Hyperion campus with plans to expand from 2 GW to 5 GW, while Microsoft and Google continue building multi-gigawatt data center campuses across the US. These projects require capital commitments that few companies outside the top-tier hyperscalers can match. Morgan Stanley also noted that power availability — not financing — is increasingly becoming the biggest bottleneck. Utilities face multi-year delays adding new generation and transmission capacity, stretching construction timelines and increasing project costs. McKinsey estimates cumulative AI infrastructure spending could reach trillions of dollars by 2030, while Epoch AI projects multiple frontier AI clusters exceeding 1 GW this year alone. **What Rising Costs Mean for Investors** For Nvidia, more expensive AI factories translate into higher revenue per deployment because its chips, networking products, and software remain at the center of those installations. Suppliers of high-bandwidth memory, power management systems, and liquid cooling equipment also stand to benefit as clusters become larger and more complex. Nvidia shares, trading at roughly 30x forward earnings, have already priced in much of this infrastructure buildout. The question for investors is whether the market has fully accounted for the concentration risk — that only a handful of companies can sustain this level of spending, and any pullback from Microsoft, Amazon, or Meta could ripple through the entire AI supply chain. Morgan Stanley's revised estimates suggest the AI buildout is not slowing down, but the cost of entry is creating a competitive moat that favors the biggest players and the chipmaker at the center of it all. This article is for informational purposes only and does not constitute investment advice.
Citi Research lowered its price target on Microsoft Corp. while maintaining a positive rating, telling investors to brace for intensifying artificial intelligence spending in fiscal 2027. "Investors should brace themselves for intensifying AI spending in fiscal 2027," the Citi Research analyst said, according to a report published Wednesday. Microsoft shares have fallen 20% year to date to about $392, near their 52-week low of $349.20. The stock trades at roughly 21 times forward earnings of $18.89, well below the 10-year average multiple for the software giant. The company's market capitalization has declined by hundreds of billions of dollars this year even as its AI business surpassed $37 billion in annualized revenue, up 123% from a year earlier. The price target cut reflects a growing concern on Wall Street about the cost of Microsoft's AI infrastructure buildout. Chief Executive Officer Satya Nadella has guided to roughly $190 billion in capital expenditures for calendar year 2026, up from about $110 billion in 2025. The spending surge has compressed free cash flow to $15.8 billion in the most recent quarter, down 10% from a year earlier, even as Microsoft Cloud revenue reached $54 billion, up 29%. The rating remains positive, with Citi joining 54 other analysts who rate the stock a buy or strong buy, according to consensus data. The average analyst price target stands at $559.86, implying about 43% upside from current levels. Microsoft's commercial remaining performance obligations nearly doubled to $627 billion, suggesting customer demand can absorb the spending increase over time. The lowered target puts additional pressure on Microsoft's upcoming earnings report, expected in late July. Investors will watch whether Azure can sustain its 39% to 40% growth rate and whether Copilot seat additions — which rose 250% year over year — are converting into durable revenue. The stock's next catalyst is the fiscal fourth-quarter earnings call, where management is expected to provide updated fiscal 2027 guidance. This article is for informational purposes only and does not constitute investment advice.
Microsoft Corp. shares dropped more than 2 percent in pre-market trading on July 14, extending a 20 percent year-to-date decline, as investors weighed the software maker's $190 billion AI infrastructure spending plan against emerging threats from enterprise customers building in-house software alternatives. "The vast majority of our capex funds short-lived assets that correlate directly with revenue," Amy Hood, Microsoft's chief financial officer, said during the company's most recent earnings call, defending the spending plan that has become the central point of contention for investors. The pre-market decline pushed Microsoft shares toward $376, adding to losses in a stock that has already shed more than a fifth of its value this year. The company's AI business is running at an annualized $37 billion, up 123 percent from a year earlier, while commercial remaining performance obligations stand at $627 billion, nearly double the prior year. Capital spending reached $30.88 billion in the fiscal third quarter, up 84 percent, with management guiding to roughly $190 billion in calendar 2026. Insider activity across 33 recent transactions shows net selling, reflecting internal caution. The selloff reflects a concern that extends beyond Microsoft's own spending. Starbucks Corp. said it is developing internal AI tools to replace vendor software from Microsoft and International Business Machines Corp., targeting a $400 million annual technology budget. The coffee chain plans to cut $30 million from near-term spending, including an immediate $10 million reduction in software costs, with initial deployments replacing Microsoft inventory management systems slated for late 2027. If a non-tech company can successfully eliminate hundreds of millions in vendor spending using AI, other Fortune 500 enterprises may follow, threatening the subscription-based revenue models that underpin Microsoft's valuation. **The Capex Conundrum** Microsoft trades at 23 times trailing earnings and 20 times forward earnings, a discount to its historical average despite 23 percent quarterly earnings growth and 18 percent revenue expansion. Azure revenue grew 40 percent in the most recent quarter, and Microsoft Cloud generated $54.5 billion, up 29 percent. But free cash flow remains the pressure point: the price-to-free-cash-flow ratio sits at 39.87, with a free cash flow yield of 2.51 percent — below the 10-year Treasury yield. Depreciation from AI hardware will weigh on margins for years, analysts have noted. The 50-day moving average of $405.31 now sits well below the 200-day average of $443.59, a bearish technical signal. **Enterprise Software's Moat Under Threat** The Starbucks initiative represents a potential structural shift in enterprise technology spending. By developing proprietary AI architecture, the coffee chain can transition software costs from operating expenses to capital expenditures, amortizing development costs over time rather than paying recurring licensing fees. Industry analysts estimate that up to 20 percent of enterprise software spending could face similar disruption as companies use AI to build custom backend solutions. For Microsoft, which counts enterprise software subscriptions among its most profitable revenue streams, the trend threatens a key growth driver at a time when the company is already under pressure to demonstrate returns on its AI investment. Of 57 analysts covering Microsoft, none rate it a Sell, and the consensus price target of $559.93 implies roughly 46 percent upside from current levels. But the stock's 23 percent decline over the trailing year — against a rising S&P 500 — suggests the market is already pricing in risks that the analyst community has yet to fully acknowledge. The next catalyst for the stock will be Microsoft's fiscal fourth-quarter earnings report, expected in late July, when investors will scrutinize whether Azure growth and AI revenue can justify the accelerating capital spending. This article is for informational purposes only and does not constitute investment advice.
AI monetization and capital spending sustainability have replaced macro tail risks as the central questions for US tech stocks, with hyperscalers projected to spend more than $750 billion on AI infrastructure in 2026, according to Intellectia.ai. "Market understanding of AI is returning to rationality, with more urgent demands for short-term ROI," CITIC Securities said in a July 14 research report. The brokerage identified AI monetization, hyperscaler capex sustainability, and value chain allocation as the three focal points for the second half. The tension is already visible in chip stocks. The PHLX Semiconductor Index has gained about 60 percent this year, yet strong earnings from Samsung and Cerebras triggered sell-offs as markets priced in perfection. Samsung projected a massive profit increase on memory strength; its stock fell after a 360 percent rally over the prior 12 months. Meta Platforms plans to sell excess computing capacity — bulls see smart monetization while bears read it as a sign of overbuying. The gap between real demand and elevated expectations will sustain volatility. Pat Gelsinger, former Intel chief now at Playground Global, calls AI demand almost unlimited, with energy as the sole real constraint. Lumentum reports its data-center components sold out five years in advance. The question is whether current stock prices reflect that trajectory or have raced too far ahead. **Energy Emerges as the Binding Constraint** Power availability now dictates the pace of AI expansion. Grids cannot scale on quarterly timelines; permitting drags for years. Nvidia-backed startups raised fresh rounds this spring to address data-center electricity shortages. Turbines, transmission lines and fuel contracts move far slower than silicon wafers. Chipmakers ship product while operators wait for the juice to run it. That mismatch explains why infinite demand collides with finite reality. Hyperscalers Microsoft, Amazon, Google, Meta and Oracle could spend more than $750 billion on AI infrastructure in 2026, per Intellectia.ai. Memory prices have soared as supply stays tight. SK Hynix and Micron posted blowout quarters. Micron's latest forecast shattered estimates on insatiable AI memory demand, Bloomberg reported in June. Yet the same reports triggered sell-offs. Investors fear any slowdown in the spending spree. **Valuation Meets Reality in Chip Stocks** The current cycle differs from the dot-com era. Companies driving this rally post actual profits. Nvidia holds more than 80 percent of the AI accelerator market. AMD, Broadcom and others grab share where they can. Still, concentration worries some observers. Market share at current levels exceeds peaks seen around 2000. Hundreds of billions in capital expenditure must still prove returns. SoftBank's Masayoshi Son dismisses bubble talk, calling it an insult. The build-out represents a generational infrastructure shift, he argues. Suppliers echo the tone. Optical components for connecting thousands of GPUs inside clusters remain critically short. Demand shows no plateau. But execution must stay flawless. Any delay in new fabrication capacity, any hiccup in advanced packaging, any revision to capex plans sends tremors. Recent trading reinforces the tension. Chip stocks tumbled in early July on renewed AI anxiety, a CNBC report detailed on July 12. SK Hynix comments about moderating AI memory expansion rippled across global markets. Nvidia, Broadcom and others gave back ground quickly. Hyperscaler spending commitments remain intact. The reaction shows how sensitive valuations have become. **The Investor Takeaway** For investors, the key question is which parts of the AI value chain offer the best risk-reward. CITIC Securities recommends focusing on hyperscaler platforms, application software with revenue growth inflecting higher, and internet names — while cautioning against high-expectation cybersecurity and infrastructure software. Hardware and semiconductor trading is likely to narrow toward short-term high-certainty segments. Nvidia trades at about 22 times forward earnings, a discount to the S&P 500. Microsoft, down about 30 percent from its all-time highs, trades at 20 times forward earnings with an AI business producing $37 billion in annual recurring revenue growing at 123 percent. Meta Platforms trades at 18 times forward earnings. These valuations reflect the market's skepticism about near-term ROI — a skepticism that CITIC Securities says will define the second half. *This article is for informational purposes only and does not constitute investment advice.*
Microsoft Corp. was sued for securities fraud after the software giant's stock lost 10 percent in January when it disclosed slowing Azure cloud growth and weaker-than-expected adoption of its Copilot AI assistant. "The complaint alleges that Microsoft misled investors about Copilot's capabilities and user adoption, which artificially inflated Azure-related revenue," Adam McCall, a partner at Bleichmar Fonti & Auld LLP, said. The lawsuit, filed in the U.S. District Court for the Western District of Washington, covers investors who bought Microsoft shares between May 1, 2025 and Jan. 28, 2026. Microsoft shares fell $48.13, or 10 percent, to $433.50 on Jan. 29 after the company reported disappointing fiscal second-quarter results. Azure growth had slowed suddenly, and Microsoft 365 Copilot had only 15 million premium customers, materially below analyst estimates. The Wall Street Journal later reported on Feb. 3 that Copilot suffered from "confusing brand positioning and interoperability problems" that frustrated users. Two law firms — Bronstein, Gewirtz & Grossman LLC and Bleichmar Fonti & Auld LLP — have announced the class action, which asserts claims under Sections 10(b) and 20(a) of the Securities Exchange Act of 1934. The complaint alleges that Microsoft consistently touted Copilot's best-in-class capabilities during the class period, even as the AI assistant suffered from severe functionality issues that caused user adoption to decline and put Azure revenue at risk. The case introduces legal and regulatory risk for Microsoft, which has invested billions into AI development. Azure has been the company's main growth driver in recent years, and any disruption to that business could affect the broader technology sector. Microsoft is a major component of the S&P 500 and Nasdaq 100 indices. Investors have until Aug. 11 to seek appointment as lead plaintiff in the case, which is captioned City of St. Clair Shores Police and Fire Retirement System, et al., No. 26-cv-02071. This article is for informational purposes only and does not constitute investment advice.
Wall Street enters mega-cap tech Q2 earnings with the lowest expectations in two years, setting up a potential beat that could fuel a rally, HSBC's Max Kettner said. "The business models of mega-cap tech companies have fundamentally changed in terms of taking on debt and being cash flow negative, but the real story is that expectations are now so low that any positive surprise will have real fuel behind it," Kettner, HSBC's chief multi-asset strategist, said on CNBC's Closing Bell Overtime on July 7. The four hyperscalers — Microsoft Corp., Meta Platforms Inc., Amazon.com Inc. and Alphabet Inc. — are expected to boost combined capital expenditures to more than $470 billion in 2026 from about $350 billion in 2025, according to analyst estimates compiled by FactSet. Microsoft's fiscal second-quarter capex consensus stands at $36.25 billion, up 60 percent from a year earlier, while Meta faces nearly 57 percent growth to over $110 billion. Amazon's capex forecast of $125 billion for 2026 is the highest among the group, and Alphabet is expected to spend more than $115 billion. A broad-based beat across mega-cap tech could reignite the Nasdaq 100 and S&P 500, which have lagged as investors priced in AI spending concerns. The S&P 500 declined about 9 percent from its January highs through late March as the Middle East conflict escalated, and the Cboe Volatility Index more than doubled during that period. The earnings season kicks off this week with reports from Apple Inc., Meta, Microsoft and Tesla Inc., followed by Alphabet and Amazon next week. Investors will scrutinize whether AI revenue growth is pacing with the unprecedented capex commitments, a question that has weighed on sentiment since late 2025 when Meta's stock had its worst day in three years after lifting its spending forecast. Microsoft's operating margin is expected to narrow to 67 percent, its tightest in three years, as the company ramps data center buildout to meet Azure AI demand. Meta's capital spending could reach $125 billion in 2026, Goldman Sachs estimates, rising to $144 billion in 2027. Amazon Web Services signed a $38 billion deal with OpenAI in November, its first contract with the ChatGPT maker, intensifying competition among cloud providers for AI workloads. The broader earnings picture supports the case for positive surprises. S&P 500 companies reported 21.2 percent earnings growth in the first quarter, with 79.6 percent beating EPS estimates and 78 percent beating revenue estimates, according to Zacks Investment Research. Full-year 2026 S&P 500 earnings are expected to grow 19.7 percent, with all 16 sectors projected to post positive growth for the first time in years. The lowered bar creates asymmetric upside for the sector, Kettner argued. If companies merely meet the reduced expectations, the relief rally could be significant. If they beat, the fuel for a sustained move higher is in place. This article is for informational purposes only and does not constitute investment advice.
Microsoft Chief Executive Officer Satya Nadella criticized AI labs for complaining about model distillation, calling it hypocritical for companies that train on publicly available data to then restrict others from doing the same. "While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation, and to reserve the right to learn from customer usage and interaction data," Nadella wrote in an X post on Sunday. Model distillation involves training a smaller, less powerful AI model using the outputs of a more advanced one. Frontier AI labs including Anthropic, OpenAI and Google DeepMind have relied on publicly available writing, images and other data to train their flagship models — a practice that has drawn lawsuits from content creators and publishers. Nadella argued that if learning flows in only one direction, "owners of the learning infrastructure make all the money while creators of the knowledge get left out." Though Nadella did not name any specific company, his comments appeared aimed at Anthropic. Earlier this year, Anthropic Chief Executive Officer Dario Amodei complained that Chinese model makers were using Claude to train their own models. Last month, the company wrote to US Senators Tim Scott and Elizabeth Warren alleging that Alibaba had carried out "the largest known distillation attack" on it to date. Anthropic said competitors can "acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost, that it would take to develop them independently." **The Data Ownership Debate** Nadella warned that companies relying on leading AI models are effectively handing over their proprietary data and then paying to use the same models. He said enterprises should own their AI infrastructure and institutional knowledge rather than depend on any single model vendor. Companies should conduct their own evaluations and maintain their own "learning loop" to improve AI capabilities continuously, he said. "That is why enterprises need a real trust boundary for their human capital and token capital to compound," Nadella wrote. "And it is a hard boundary across which nothing crosses, not even the intelligence exhaust, without consent." The debate highlights a growing tension in the AI industry over who controls the data and infrastructure that power modern models. Elon Musk has also criticized Anthropic's data practices, writing in a February X post that the company is "guilty of stealing training data at massive scale." **What It Means for Investors** For Microsoft, the comments reinforce the company's strategy of selling AI infrastructure through Azure rather than competing directly with frontier model makers. Microsoft has invested more than $13 billion in OpenAI but also offers access to models from multiple providers through its cloud platform. The tension between model providers and infrastructure owners could accelerate a shift toward enterprise-owned AI systems, reducing dependence on any single vendor. Companies building proprietary AI stacks — including Microsoft, Amazon and Google — stand to benefit as enterprises seek greater control over their data and model training pipelines. This article is for informational purposes only and does not constitute investment advice.
**Major technology companies have borrowed nearly a quarter-trillion dollars through corporate bond sales to fund artificial intelligence infrastructure, creating a supply glut that is pushing down bond prices and testing the limits of investor demand.** Tech giants have issued roughly $250 billion in corporate bonds to finance AI infrastructure this year, the largest wave of debt sales tied to a single technology theme, weighing on bond prices and stretching investor capacity. "The market is absorbing an unprecedented volume of single-name supply tied to one investment theme, and the price action in secondary trading shows clear indigestion," said Hans Mikkelsen, managing director of credit strategy at TD Securities. The issuance has pushed credit spreads wider from the tight levels seen early in the year, with the average investment-grade corporate bond spread now at 0.74 percentage points over Treasuries, near multi-decade lows but under pressure. Primary dealers have turned net short on corporate bonds for the first time on record, holding a net-short position of roughly $4 billion, concentrated in longer-dated maturities where $13.7 billion of short positions are offset by $9.66 billion of longs in shorter-dated debt, according to Crisil Coalition Greenwich data going back to 1998. The supply wave threatens to raise borrowing costs for other corporations seeking to issue debt, potentially tightening financial conditions at a time when the Federal Reserve is maintaining elevated interest rates. If investor demand falters further, the pipeline of planned AI-related bond sales could face pricing pressure or delays, with implications for the pace of data center construction and AI chip procurement. The bond sales have been concentrated among the largest US technology companies, including Microsoft, Amazon, Alphabet and Meta Platforms, each raising billions to fund data center construction, graphics processing unit purchases and electricity capacity agreements. The combined spending on AI infrastructure by the four companies is expected to exceed $200 billion this year, with bond markets serving as the primary funding channel. The supply dynamic has created an unusual market structure. Electronic execution now handles 49 percent of investment-grade trades, up from 8 percent a little over a decade ago, allowing dealers to match client flows without warehousing risk. Yet the persistent inflow of new bonds has left primary dealers unable to hold inventory, pushing them into an aggregate net-short position for the first time in the data series. Strong demand from insurers, pensions and money managers reinvesting higher-yielding coupons has provided a floor under the market. But Lazard strategists have flagged the asymmetry in positioning, warning that if credit spreads tighten further, dealers forced to cover their shorts into thin supply could amplify a rally — a scenario that would benefit existing bondholders but complicate execution for new issuers. The $250 billion figure represents issuance across investment-grade and high-yield markets, with the bulk coming from companies with strong credit ratings. The wave has drawn comparisons to the surge in energy-sector bond issuance during the US shale boom, though the scale of AI-related debt sales has already surpassed that period in dollar terms. For investors, the key question is whether demand from yield-seeking institutional buyers can keep pace with supply. The answer will shape borrowing costs for the technology sector and determine how much of the AI build-out is funded through debt rather than cash flow or equity issuance. This article is for informational purposes only and does not constitute investment advice.
The cost of powering artificial intelligence is reshaping industrial equipment markets. Gas turbine prices have surged roughly 300% over three years, Melius Research estimates, as data center operators compete for limited generation equipment. "Power is the lifeblood of your business. If they don't have power, their business doesn't exist," Jeremy Eliahou Ontiveros, an energy analyst, said on the *Catalyst with Shayle Kann* podcast. He warned that AI labs like Anthropic are trying to scale from 1.5 gigawatts of capacity at the end of 2025 to more than 10 gigawatts by 2027 — the equivalent of building a Google-sized power footprint in two years. Microsoft recently ordered seven large gas turbines from GE Vernova for a Texas data center, with each unit costing more than $250 million. GE Vernova's turbine supply is effectively booked through 2030, and its Electrification segment booked $2.4 billion in data center equipment orders in the first quarter of 2026 — more than all of last year. The company's shares have gained 64.8% year to date. The supply crunch reflects a structural gap. Goldman Sachs projects US data center electricity consumption will double between 2025 and 2027, while the grid can only add roughly 35 gigawatts of new supply annually, according to analyst estimates. That leaves a 15-gigawatt annual shortfall against the 50 gigawatts of data center demand the industry is targeting. Research outfit RAND expects "behind the meter" power generation — facilities that bypass the grid entirely — to roughly triple to 49 gigawatts by 2030. **The Bypass Model Gains Traction** Chevron is capitalizing on the trend through a novel structure. The oil and gas company partnered with GE Vernova to supply Microsoft's West Texas data center under a 20-year agreement, bypassing traditional utility providers. Chevron supplies the natural gas while GE Vernova provides the turbines — a self-contained power solution that could become the norm for AI data centers, according to analysts. PwC estimates AI-linked natural gas demand could quintuple between 2025 and 2035. The model addresses a bottleneck that extends beyond generation. Host Shayle Kann noted on the podcast that even if new generation is built, upgrading substations and ordering high-voltage transformers takes three years. That timeline mismatch between grid infrastructure and AI buildout schedules is pushing hyperscalers toward self-generation. **Who Wins, Who Loses** The equipment suppliers are the clearest beneficiaries. GE Vernova's turbine backlog and slot reservation agreements are on track to reach at least 110 gigawatts by year-end 2026, CEO Scott Strazik said. Vertiv, a pure-play data center power and cooling company, reported organic orders up 252% year over year in the fourth quarter of 2025, with backlog reaching $15 billion. Eaton's Electrical Americas orders rose 42% organic on a trailing-twelve-month basis, and the company closed a $9.55 billion acquisition to expand data center cooling capacity. On the generation side, Constellation Energy closed its Calpine acquisition in January 2026, operating a 55-gigawatt fleet as the largest private power producer globally. Its first-quarter revenue jumped 63.9% year over year to $11.12 billion. Vistra signed 20-year power purchase agreements with Meta for more than 2,600 megawatts and with Amazon Web Services for up to 1,200 megawatts. GE Vernova shares trade near analyst targets averaging $1,222.63, while Vertiv has doubled year to date. The open question is whether the roughly 35 gigawatts of annual new supply can meet the 50 gigawatts of annual data center demand — and whether the transformers, substations, and turbines arrive in time. For investors, the equipment and generation layers of the AI trade offer the tightest supply and the loudest demand signal. This article is for informational purposes only and does not constitute investment advice.
**More than $200 billion in data-center debt was raised in 2025, creating a closed-loop financing system with echoes of past bubbles.** More than $200 billion in data-center debt was raised in 2025, with projections topping $1 trillion by 2028, as cross-investments among Nvidia, Microsoft and OpenAI created financing structures one analyst likened to a blend of 1990s infrastructure spending and 1920s-style lending. "AI is real, but investors may be pricing the future far too early," Ray Dalio, founder of Bridgewater Associates, said. Nvidia passed a $5 trillion valuation in October 2025 and is pouring $100 billion into OpenAI to help build data centers loaded with its chips. Microsoft owns 27 percent of OpenAI and represents nearly a fifth of Nvidia's revenue. OpenAI partners with CoreWeave, a company Nvidia also holds a large stake in. When CoreWeave issues billions in debt to build new capacity, Nvidia guarantees it will buy whatever CoreWeave cannot sell through 2032. The risk is that the entire structure depends on constant capital inflows. OpenAI, valued at $500 billion, expects $13 billion in revenue and a $5 billion loss in 2025 and may burn more than $140 billion before turning profitable — more than Amazon, Tesla and Uber's cumulative early losses combined. If credit conditions tighten or a major player needs to sell instead of buy, the system could face a rolling reset similar to 2000-2002. **A $1 Trillion Debt Market Takes Shape** Roughly $200 billion in data-center debt was raised in 2025 alone. By 2028, the market could exceed $1 trillion, with as much as $750 billion coming from private credit, according to industry estimates. Meta secured $29 billion for a Louisiana facility, with $26 billion structured as debt from Blue Owl and PIMCO. Many of these deals involve off-balance-sheet financing, GPU-backed collateral and complicated leasing structures with borrowers who have never been tested in a downturn. AXA said in December it would "avoid financing technological gambles" after watching lending volumes explode, a sign that even major institutions are starting to worry. Wisconsin regulators in July denied We Energies' petition to loosen financial guarantees for hyperscale data center users, suggesting oversight is tightening. **The Revenue Gap That Could Break the Loop** The most awkward reality: the companies building the foundation for AI are not profitable. OpenAI may burn more than $140 billion before turning profitable — more than Amazon, Tesla and Uber's cumulative early losses combined. An MIT study found 95 percent of companies see zero return on their generative-AI investments despite spending $30 billion to $40 billion. Bain estimates that AI will need $2 trillion in annual revenue by 2030 just to justify current infrastructure spending — more than the combined revenues of America's largest tech firms in 2024. Goldman Sachs notes that $19 trillion in market cap is running ahead of economic impact, citing five danger signals reminiscent of the 1990s: peaking investment, falling profits, rising debt, Fed cuts and widening credit spreads. The S&P tech sector trades at 30 times forward earnings — high, but nowhere near dot-com extremes. Still, AI has driven 75 percent of S&P returns, 80 percent of earnings growth and 90 percent of capex growth since late 2022, making the market increasingly dependent on a handful of companies. The most likely scenario is not a sudden collapse but a rolling reset where the weakest players fail first and drag parts of the system with them. Big Tech will survive. The casualties will be the unicorns, the developers relying on off-balance-sheet structures and the investors who believed the hype without checking the math. Data-center demand is projected to grow more than 19 percent annually through 2030, and Nvidia sees global capex rising from $600 billion to as much as $4 trillion. If demand keeps up, today's spending rush will not look reckless — it will look early. The question for 2026 is whether the financial structure built around it can hold. This article is for informational purposes only and does not constitute investment advice.
Chevron is moving beyond traditional oil and gas supply, partnering with GE Vernova to provide natural gas for a Microsoft AI data center in West Texas under a 20-year agreement that bypasses conventional utility grids. GE Vernova will supply the natural gas turbines while Chevron provides the fuel, creating a self-contained power solution for one of Microsoft's most energy-intensive facilities. "The offers that you get for using the compute are so high that it may make sense, in some cases, to rent out or consider those kind of deals instead of your own internal uses," Meta Chief Executive Officer Mark Zuckerberg told Bloomberg News, reflecting the broader industry shift toward self-supplied power. U.S. data center electricity consumption is projected to double between 2025 and 2027, according to Goldman Sachs research. Behind-the-meter power generation capacity is expected to roughly triple by 2030 to 49 gigawatts, RAND estimates show. AI-linked demand for natural gas could more than quintuple between 2024 and 2035, according to a PwC outlook. WTI crude traded near $76 a barrel last week as energy stocks rallied on renewed Middle East tensions, reflecting the broader supply dynamics shaping the sector. For Chevron, the agreement opens a new revenue stream that uses existing gas infrastructure without requiring costly new construction. For the AI industry, it shows that the nation's power grids and utilities cannot keep pace with the energy demands of data centers, pushing the world's largest technology companies to build their own power solutions. **Behind the Meter: A $49 GW Opportunity** The deal structure represents a fundamental shift in how large energy consumers procure power. Rather than waiting for utilities to expand grid capacity — a process that can take years and requires regulatory approval — technology companies are contracting directly with fuel suppliers and turbine manufacturers. Chevron's existing natural gas collection and pipeline network in West Texas allows it to serve this market without the capital expenditure typically required for new energy infrastructure. The opportunity extends beyond Chevron. Research outfit RAND expects the nation's behind-the-meter power generation capacity to roughly triple between now and 2030, reaching 49 gigawatts. Natural gas turbines are expected to be the single largest source of this capacity expansion, benefiting companies with existing infrastructure that can be deployed quickly. **The Emissions Trade-Off** The environmental cost of this approach is significant. Permits for the Chevron-powered Microsoft data center in West Texas show the facility could emit more than 11.5 million tons of CO2 equivalent annually — an amount greater than the entire state of Rhode Island's annual emissions. Microsoft's greenhouse gas pollution increased roughly 25 percent last year, driven primarily by data center expansion, according to the company's latest sustainability report. Microsoft said it matched 100 percent of its electricity consumption with carbon-free sources in the reporting period and still plans to become carbon negative by 2030. The company has also stopped purchasing unbundled renewable energy certificates, a move that contributed to the rise in its Scope 2 emissions but that researchers described as commendable for prioritizing investments in new clean electricity. States with stronger climate policies are pushing back. New York legislation would require data centers over a certain size to meet renewable energy benchmarks starting in 2030 and get at least 90 percent of their energy from renewable sources by 2040. Michigan, Oregon and Minnesota have enacted laws designed to protect their existing requirements that electric utilities use only emissions-free energy sources by 2040. This article is for informational purposes only and does not constitute investment advice.