

A 3.5-million-barrel miss against consensus landed on the wrong side of positioning last week, as the U.S. Energy Information Administration reported Wednesday that commercial crude stockpiles grew by 3 million barrels instead of shrinking by the 500,000 barrels analysts had penciled in. The build lifted commercial inventories to 426.4 million barrels, putting them 2 percent above the five-year average for this time of year, according to the EIA's weekly petroleum status report. The American Petroleum Institute had flagged the direction a day earlier with a smaller 1.786-million-barrel gain. "The headline number is bearish, but the barrel count is doing all the work here," said Omar Tariq, commodities strategist covering oil and gas markets. "Crude built because imports ran and refiners were still working through maintenance season. The product side tells a different story." Gasoline inventories fell 1.7 million barrels after a 800,000-barrel gain the prior week, and middle distillates — the diesel and heating oil pool — drew down 400,000 barrels. Distillate stocks now sit 12 percent below the five-year average, a gap that has persisted through the third quarter and leaves the heating season with thin cover. The demand read is where the bearish case weakens. Total products supplied, the EIA's proxy for U.S. oil consumption, averaged 20.6 million barrels per day over the four weeks to September 18, up 0.5 percent from the same stretch a year earlier. Gasoline demand held at 8.8 million barrels per day on the same four-week basis, while distillate supplied averaged 3.6 million barrels per day, down 0.3 percent year over year. Prices moved against the inventory signal. Brent crude traded at $101.40 a barrel at 10:05 a.m. in New York, up $2.15, or 2.17 percent, on the day, though still about $6 below where it sat a week earlier. WTI gained $1.47, or 1.63 percent, to $92.00, roughly $10 lower than the prior week. The divergence between a bearish stockpile print and a higher settle points to a market trading macro risk appetite over a single weekly data point. The last time crude stocks posted a build of this size against a consensus draw, in the week ended April 24, WTI fell 3.1 percent over the following five sessions before recovering as refinery runs climbed back above 90 percent of capacity. That pattern matters for how traders treat this print: builds driven by import timing and maintenance tend to reverse within two to three reporting cycles, while builds driven by genuine demand erosion do not. For the energy sector, the immediate read is a modest drag on integrated majors and shale producers whose cash flow assumptions lean on WTI holding above $90. A sustained move below that level would pressure buyback capacity at the large-cap producers and complicate fourth-quarter capital budgets. The next test arrives with the EIA's report for the week ended September 25, due October 1, which will show whether refinery utilization has recovered enough to pull crude out of the system. This article is for informational purposes only and does not constitute investment advice.

Alibaba's cloud unit grew external revenue 45% in the June quarter, a 22-quarter high, while capital expenditure climbed 75% year over year to RMB 67.68 billion ($9.5 billion) — a spending pace that cut quarterly net profit by 75% and left investors weighing whether the AI buildout pays for itself. "AI-related products now represent 35% of external cloud revenue after 12 consecutive quarters of triple-digit growth," Alibaba management said on the June-quarter earnings call, guiding the annualized AI run rate toward $10 billion in the following quarter. The arithmetic behind that guidance is where the tension sits. Alibaba Cloud's external revenue growth accelerated to 45% from 3% headline growth in fiscal 2026, and the segment's EBITDA margin already stands at 12% and is expanding. Management estimates a 2.5-year payback on AI infrastructure, which would place the earnings recovery inside the next 18 months. Against that, operating margin fell to 6% of revenue from 14% a year earlier, and the company raised $10.2 billion in August through 710 million new shares at HK$112.70 — a 3.6% dilution at an 8.4% discount to the prior close. Bernstein analyst Robin Zhu cut his price target to $165 from $180 while keeping a Buy rating, citing execution risk and the credibility cost of raising capital after the company had highlighted a strong cash position. Susquehanna took the other side, maintaining a Positive rating and lifting its target to $190 in late August. The competitive frame is what makes the capex number legible. Alibaba's Qwen family competes directly with DeepSeek's models on cost-per-token economics and with ByteDance's Doubao on domestic enterprise adoption, while its in-house AI chips are designed to reduce reliance on Nvidia's H20 and Huawei's Ascend 910B for inference workloads. Every dollar of that RMB 67.68 billion quarterly capex is a bet that owning the silicon and the model stack lowers the marginal cost of serving tokens below what a pure cloud reseller can match. That bet carries a policy dimension Alibaba does not control. U.S. export controls have repeatedly reshaped which accelerators Chinese cloud providers can buy, pushing Alibaba toward in-house designs and domestic foundry capacity. The company has not disclosed the process node or performance specifications of its latest AI chip, and did not disclose the test conditions for any internal cost comparisons against Nvidia or Huawei parts. The stock reflects the disagreement. BABA's tokenized shares traded at $115.40 on Sept. 21, above the 7-day simple moving average of $112.25 and the 20-day at $111.50 but below the 50-day at $118.81 — a recovery, not a resumption. The forward P/E of 17.67 sits against a consensus mean target of roughly $184 to $188 across 26 covering analysts, 21 of whom rate the stock a Strong Buy. The PEG ratio of 0.51 prices hyper-growth cloud metrics like a slow-growth value name. The dilution overhang has not fully cleared, and a securities class action carries an Oct. 5 lead-plaintiff deadline. Offsetting that, Chairman Joe Tsai and CEO Eddie Wu bought a combined HK$120 million of BABA shares in the open market after the equity raise, and the company holds $385.7 billion in total cash. The next verdict point is Alibaba's November earnings report, estimated for Nov. 24. If cloud external revenue holds above 45% growth and segment EBITDA margin expands, the forward-multiple de-rating case weakens and a move toward $135 to $140 becomes the base case on a simple re-rating to 22-23x. If capex keeps outrunning monetization, the 75% net-profit decline stops looking like a buildout phase and starts looking like the run rate. This article is for informational purposes only and does not constitute investment advice.

The 10-year Treasury yield climbed above 5% for the first time in this cycle, a move that raises the risk-free discount rate applied to every long-duration asset and pushed the Nasdaq Composite down 0.95% to 26,986.44. The benchmark yield settled at 5.08%, up 2.21% on the session, according to Seeking Alpha data, while the 2-year note — the maturity most sensitive to Federal Reserve policy — rose 2.82% to 4.89%. The 30-year reached 5.38% and the 5-year 4.96%, leaving the entire curve above 4.8%. The repricing follows the Fed's decision to raise its benchmark rate by 25 basis points to a target range of 3.75%-4%, the first increase since July 2023. Chair Kevin Warsh told reporters that inflation has been "too high ... for too long," adding that the Federal Open Market Committee had decided "this standard has not been satisfied." The dot plot — the individual rate projections submitted by each Fed official — showed 16 of the 18 participants expecting at least one more increase this year, with four seeing two. That guidance, rather than the hike itself, is what lifted the long end: traders who had positioned for a pause into year-end were forced to reprice the terminal rate higher. "The bond market's biggest moves are likely now in the rearview mirror," said Bob Edwards, chief investment officer at Florida-based Edwards Asset Management, in a note. "There is now a good opportunity for investors after this big move to lock-in these elevated yields." Edwards added that any further increase would most likely arrive at the December meeting, since the Fed is unlikely to move at its October gathering, which falls days before the midterm elections. The move was not linear. Yields slid on Thursday, with the 10-year falling more than 7 basis points to 4.93% and the 2-year slipping more than 5 basis points to 4.67%, before Friday's rebound erased that retreat. Weekly jobless claims for the period ended Sept. 12 came in at 196,000, below the 207,000 economists polled by Dow Jones had expected — a reading that argued against an imminent policy pivot and helped cap the bond bid. ## Duration Is the Transmission Channel For equities, the arithmetic is mechanical. A higher risk-free rate lowers the present value of every future cash flow, and the effect scales with how far out those cash flows sit. That is why the Nasdaq Composite's 0.95% decline outpaced the broader market: software, semiconductor and artificial intelligence names carry the longest duration in the index, and their valuations rest on earnings projected years into the future. The same discount-rate logic reaches Bitcoin, which trades as the longest-duration asset in the market — it produces no cash flow at all, so its price is entirely a function of the rate used to discount the future. Rising Treasury yields pull capital toward fixed income by making a risk-free 5.08% return available, and they tighten the financial conditions that speculative assets depend on. The pressure is a function of the yield level, not of anything specific to crypto. The cross-asset chain runs in one direction: Fed guidance lifts the front end, the front end drags the long end higher, the long end compresses equity multiples, and the same rate move draws capital out of speculative assets. Each link is a repricing of the same variable. ## What the Next Fed Meeting Decides The immediate question is whether 5.08% holds. The 10-year has now retraced a pullback to 4.93% within two sessions, and the speed of that reversal suggests sellers remain in control of the long end. A sustained break above 5.1% would put the next leg of equity multiple compression in play; a failure to hold would mark the new high as a spike rather than a regime change. The catalyst that resolves it is the Fed's next meeting. With 16 of 18 officials projecting another increase and four open to two, the burden of proof sits with the data. Warsh's framing — that the committee needs confidence inflation is "moving to our objective clearly and at sufficient speed" — leaves little room for a pause unless the incoming prints soften materially. Political pressure adds a variable the bond market has historically priced with a discount. President Donald Trump said in a social media post that rates "should be 1 per cent, or less, because we are the best credit in the world," and told reporters the Fed's board is "very hostile ... very political" and "doing the wrong thing." The last time a sitting president campaigned this openly against Fed policy, the market's response was measured in term premium rather than in the policy path — a distinction that matters for anyone holding duration. For portfolio positioning, the practical implication is that the discount rate is now the dominant input across asset classes, and it is set by a committee that has signaled it is not finished. Tech exposure and Bitcoin exposure are, for now, the same trade expressed twice. This article is for informational purposes only and does not constitute investment advice.