IMPACT Therapeutics signs EUR423.5 million senaparib licensing deal with Pharmanovia

source:

[1] IMPACT THERAP-B (07630.HK) Signs Exclusive Licensing Agreement for Senaparib; Shares Rise Over 10% (https://www.aastocks.com/en/stocks/news/aafn- ...)[2] IMPACT Therapeutics Strikes EUR423.5 Million Global Senaparib Deal With Pharmanovia - TipRanks (https://www.tipranks.com/news/company-announc ...)

Related News

Jul 31 2026, 04:52

Nvidia's Huang says memory is AI's biggest bottleneck, $48.6B cash ready

Nvidia's Jensen Huang said memory chips are now AI's biggest bottleneck, as the company shifts from selling standalone GPUs to building systems that pack high-bandwidth memory (HBM) directly into its hardware. Huang highlighted the constraint in recent remarks, pointing to the specialized memory chips now integral to Nvidia's AI systems. The company relies on a small group of suppliers — Micron, SK Hynix, and Samsung — for continuous high-volume HBM supply, and manufacturing those chips is far from simple. Nvidia generated $48.6 billion in free cash flow in the quarter ended April 26 and finished with $13.2 billion in cash and cash equivalents, giving it the purchasing scale to secure priority access to HBM supply. The company's stock is up 0.60 percent year to date as of market close on July 29. The memory shortage puts Nvidia at the mercy of its suppliers, but its cash position lets it pay premiums to corner the market, potentially shutting out smaller competitors and reinforcing its dominance in AI infrastructure. **Why Memory Became the Constraint** At the beginning of the current AI boom, tech giants focused on acquiring as much compute power as possible. Now the emphasis has shifted to memory chips, which have become the limiting factor in scaling AI systems. AI training and inference rely on trillions of data points, and storing and retrieving them quickly requires specialized memory chips. As AI handles more complex tasks — running autonomous agents or processing complicated context rather than generating recipes or travel recommendations — the need for HBM has grown sharply. HBM stacks multiple DRAM dies vertically and connects them through a high-speed interface, delivering far greater bandwidth than conventional memory. This architecture is essential for feeding data to GPUs at the speeds AI workloads demand. But producing HBM requires advanced manufacturing processes and specialized packaging, which is why only a handful of companies — SK Hynix, Samsung, and Micron — control the vast majority of supply. If suppliers can't produce HBM fast enough, Nvidia's own production timeline faces delays, potentially affecting its business. The constraint is particularly acute as AI models grow larger and require more memory per accelerator. **Nvidia's Cash Advantage** The upside for Nvidia is that its scale gives it first dibs on available memory supply. With $48.6 billion in free cash flow and $13.2 billion in cash on hand, the company can easily pay a premium to buy HBM in bulk, shutting out smaller competitors and further cementing its stronghold on the industry. Investors shouldn't read Huang's message as a warning sign. It's the next chapter of the AI evolution — one where Nvidia uses its leadership position to be a long-term force regardless of the current hiccup. The stock's flat year-to-date performance reflects broader sentiment around big tech and the "Magnificent Seven" stocks rather than disappointing business performance from Nvidia. For memory chipmakers, Huang's comments point to increased demand and pricing power. SK Hynix, Samsung, and Micron stand to benefit from Nvidia's willingness to pay premiums for HBM supply, potentially supporting their own revenue growth as AI infrastructure spending continues to expand. Nvidia's pivot also raises the stakes for TSMC, which handles advanced packaging for HBM integration, and for AMD, which competes with Nvidia in AI accelerators and faces the same memory supply constraints. The competitive dynamic is shifting. Nvidia's ability to outbid rivals for HBM supply could widen its lead in AI accelerators, where AMD and custom silicon from cloud providers are trying to gain traction. But the memory constraint also creates an opening: if Nvidia can't secure enough HBM, competitors with alternative memory strategies could gain ground. For investors, the key question is whether Nvidia's cash advantage translates into sustained supply security. The company's $48.6 billion in quarterly free cash flow gives it enormous purchasing power, but HBM production capacity is finite and expanding slowly. Memory makers are investing heavily to increase output, yet the timeline for meaningful capacity additions remains uncertain. This article is for informational purposes only and does not constitute investment advice.

Jul 31 2026, 04:48

Brent slips as Hormuz traffic recovers to 35% of pre-war levels

Oil prices fell Friday as Commonwealth Bank of Australia estimated crude flows through the Strait of Hormuz have recovered to roughly 30%-35% of pre-war levels, draining the geopolitical premium that pushed Brent above $100 for the first time since May. "At around $100, Brent is trading just $13 above our estimated fair value for July of $87 per barrel, implying that today's market appears to be pricing in only a modest geopolitical premium," said Natasha Kaneva, Head of Global Commodities Strategy at J.P. Morgan. J.P. Morgan analysts said commercial traffic through the Strait is running at 50 percent of pre-war levels, or 11.1 million barrels per day, including nearly 7.0 million barrels per day of pipeline re-routing. Global oil demand fell by roughly 5.1 million barrels per day, offsetting nearly 46 percent of the supply loss, while China alone cut crude imports by about five million barrels per day. The path of prices now hinges on how long disruptions persist. J.P. Morgan projects Brent will average around $94 per month if the conflict is contained to one month, but each additional month of disruption adds roughly $7-8 per barrel, lifting monthly averages to about $114 if disruptions extend to three months. The recovery in Hormuz traffic marks a turning point after a conflict that propelled Brent up nearly 40 percent in July, putting it on pace for its strongest monthly gain since March. Brent closed at $100.69 per barrel on July 23, the first time it settled at or above $100 since May. The last time Brent traded above that threshold was before the conflict disrupted flows through the vital waterway that carries roughly one-fifth of global oil consumption. Standard Chartered Bank's Energy Research Head Emily Ashford said price moves will remain "headline-driven, taking direction from near-term escalation and de-escalation in the U.S.-Iran conflict." The bank maintained its third-quarter Brent average forecast of $85 per barrel "on the expectation of sustained geopolitical risk, ongoing supply disruption, and dwindling strategic reserves." Ashford also flagged that the fragility of the memorandum of understanding and ceasefire, along with renewed attacks on vessels, has paused the normalization process. **The demand adjustment that capped prices** The demand-side adjustment has been extraordinary. J.P. Morgan analysts said the bulk of the demand loss was driven by physical shortages rather than higher prices, calling it "a forced correction, not the traditional form of demand destruction." OECD countries accounted for around 18 percent of the global demand loss, with the burden falling disproportionately on non-OECD countries and China. China's ability to sustain crude imports roughly four million barrels per day below normal for another three months provides an ongoing buffer for the global market, according to J.P. Morgan. But the analysts warned that eroding global inventory buffers are finite. The world would need to release an additional 210 million barrels if the conflict lasts one month, 315 million barrels if it lasts two months, and nearly 500 million barrels if it extends to three months, on top of the 500 million already drawn down since the conflict began. The U.S., which shouldered most of the inventory releases in April and May, is unlikely to repeat that performance, J.P. Morgan said, leaving Europe, Japan, and South Korea to replace the fading U.S. inventory impulse or allow demand to crater. Pipeline re-routing flows are becoming increasingly vulnerable to severe disruptions following reports that the Houthis have begun enforcing a Red Sea blockade, the analysts noted. If flows through Hormuz and the Red Sea fall by another four million barrels per day from current levels, the market could still avoid an acute shortage, but only if demand stays close to today's depressed levels. China cuts both ways, J.P. Morgan noted: the same import restraint that helped cap prices becomes a tailwind once it reverses, with even a one million barrel per day recovery in imports from September adding about $3 to the fair value. This article is for informational purposes only and does not constitute investment advice.

Jul 31 2026, 04:22

CoreWeave CDS hits 50% default odds as AI debt spiral tightens

CoreWeave's credit default swap spread hit 855 basis points on July 29, implying a 50 percent five-year default probability — the bond market's starkest verdict on the AI infrastructure debt model. The company's shares fell another 9 percent to close around $61.53, down more than 40 percent from a year ago, as the signal rippled through the neocloud sector. "There is a reason Nvidia's five-year credit default swaps are going parabolic," investor Michael Burry, who shorted mortgage-backed securities before the 2008 financial crisis, posted on X on July 28, attributing the move to circular spending patterns in which Nvidia invests in GPU cloud companies that use the capital to buy Nvidia's own hardware. The CDS signal arrived as a CITIC Construction Investment macro report published the same day quantified the structural gap: Microsoft, Google, Amazon, Meta, and Oracle's combined 2026 capital expenditure of roughly $780 billion exceeds their projected operating cash flow of about $650 billion, creating a $130 billion funding hole. CoreWeave's own fixed payment wall — debt and lease obligations due through 2026 — stands at $9.2 billion against expected operating cash flow of $6.2 billion, a 149 percent coverage gap that requires continuous refinancing. The divergence between credit and equity markets is stark: twenty-one analysts carry an average price target of $145.76 on CoreWeave, roughly 137 percent above Tuesday's close, while the CDS market prices a coin flip on survival. CoreWeave's August 11 second-quarter earnings call will determine which framework is correctly calibrated. **The Debt Treadmill That Powers the AI Buildout** The financing structure underpinning the AI supercycle has evolved from tech giants' internal cash flows into a credit expansion cycle that capitalizes unearned AI revenue through corporate credit, project assets, and long-term contracts. CITIC's report traces the money path: household and national savings flow through pension funds, insurers, and sovereign wealth funds into infrastructure funds, private credit vehicles, and asset managers, which in turn fund AI projects through equity, project debt, corporate bonds, and asset-backed securities. The five largest hyperscalers have shifted from balance-sheet funding to a three-layer structure of parent-company debt, long-term leases, and project financing. Moody's counts roughly $969 billion in lease commitments across the top five U.S. tech companies, most tied to data centers not yet delivered. CoreWeave's $25.1 billion interest-bearing debt stack as of March 2026 spans delayed-draw term loans, equipment financing, high-yield bonds, convertibles, and revolving credit — a capital structure that prices risk at dramatically different levels depending on the collateral. Its senior unsecured debt costs about 10 percent, while the DDTL 4.0 facility backed by investment-grade customer contracts and GPU assets prices at roughly 5.9 percent. The cost gradient reflects the market's assessment of cash flow certainty. Meta's 2032 corporate bonds carry a 4.6 percent coupon, while Hyperion's data center project bonds yield 6.58 percent — project debt is not cheaper than parent debt, it exists to preserve parent balance sheets and rating headroom. **Why CoreWeave Is the Canary** CITIC's report identifies AI cloud service providers — CoreWeave and Nebius among them — as the most dangerous liquidity risk points in 2026-2027. CoreWeave's mismatch is the most acute: its 2026 fixed payment wall of $9.2 billion equals 149 percent of expected operating cash flow, rising to roughly 170 percent when construction-linked leases and equipment purchase commitments are included. Even under a more generous 20-year amortization assumption, the ratio stays above 100 percent. The company's first-quarter numbers illustrate the treadmill. CoreWeave spent $6.8 billion on capital expenditures while generating $2.98 billion in operating cash flow, producing negative free cash flow of $4.71 billion for the quarter. First-quarter interest expense of $536 million consumed 46 percent of adjusted EBITDA. Management guided second-quarter interest expense to $650-730 million. The bear case extends beyond the income statement. Meta's July 17 announcement of Meta Compute — a commercial cloud service selling bare GPU capacity — sent CoreWeave shares down roughly 35 percent over the following days, even though Rosenblatt Securities analyst John McPeake argues the $21 billion Meta contract through 2032 explicitly prevents resale of CoreWeave's contracted capacity. Oracle's CDS has climbed above 215 basis points from roughly 145 at the end of 2025, and S&P cut Oracle to BBB-, the lowest investment grade, earlier in July. Moody's separately warned that unprecedented AI spending threatens the credit quality of Microsoft, Amazon, Alphabet, Meta, Oracle, and CoreWeave — six companies carrying combined direct debt of roughly $460 billion and $1.2 trillion in lease commitments. The August 11 earnings call must answer three questions: whether second-quarter interest expense landed within guidance, whether revenue reached the $2.53 billion midpoint, and whether management can articulate a credible path to positive free cash flow. The $99.4 billion revenue backlog — CoreWeave's central argument that the debt structure is serviceable — predates Meta's cloud announcement and may no longer reflect the competitive reality. For investors, the stakes extend beyond CoreWeave. The AI-linked credit default swap market traded approximately $650 million in the second quarter, up 600 percent year over year. If the bond market's assessment proves correct, the $4.51 trillion global AI infrastructure investment cycle through 2030 — of which tech giants contribute $2.93 trillion and AI cloud providers $248 billion — faces a repricing that would cascade through pension funds, insurers, and index funds holding CoreWeave debt and equity alike. This article is for informational purposes only and does not constitute investment advice.