Adani and Jabil plan 50:50 AI hardware venture in India as chip-equipment stocks slide on profit-taking
Digi Power X secured a $200 million Goldman-led debt commitment for its Alabama AI data center, and a tally put 2026 funding for US AI data center startups at $33.5 billion. Microsoft is weighing cost cuts in its Azure unit, while residents near Seneca Lake and in a New York town facing a 1-gigawatt project pushed back.
News +0.51Market no trading session in this window: the score is the news reading
Key notes
50%
Adani stake in Jabil hardware venture
Adani and Jabil plan a joint venture to make AI data center equipment in India. Reuters reported the two signed a pact for the venture, and marketscreener and digitimes described it as a 50:50 split.
Digi Power X lined up debt to finish its Alabama AI data center. Dealroom reported the $200 million deal is Goldman-led, and Stock Titan noted that closing depends on lender checks.
Semiconductor equipment shares fell as investors took profits. 24/7 Wall St. reported Applied Materials sank 7% and Lam Research and KLA dropped 4%, while GuruFocus reported ASML raised its 2026 revenue outlook.
Capital continues to flow to AI data center builders. ET Datacenters put 2026 funding for US AI data center startups at $33.5 billion, and Seeking Alpha reported CoreWeave is seeking insurers' capital to finance its build-out.
Chipmakers set out long-range growth claims. Yonhap reported SK hynix projects AI infrastructure performance to grow 20-fold by 2030, and Engineer Live reported the semiconductor sector hit a $1 trillion milestone as SEMICON West 2026 opened.
Local resistance to AI data centers grew on cost and power concerns. The Times of India reported a New York town once known for IBM is resisting a 1-gigawatt AI data centre, and Futurism reported the Federal Reserve says the buildout is making everything more expensive.
Data centers · 55 storiesSites, colocation, cooling, capacity leases.Digi Power X lines up Alabama debt as local opposition spreadsFinancing and new sites moved forward in the US and Morocco. Residents in New York, New Jersey, Florida and Maine pushed back against proposals.Frontier AI sites: power against capital costEach point is a site in Epoch AI's frontier register with both a stated power capacity and a stated capital cost — the further above the trend, the more each megawatt cost to build.Digi Power X announced a $200 million debt financing commitment to support completion of its Alabama AI data center.TradingViewProactive financial newsFiveNines Group was appointed to develop a 20 MW AI data centre near Casablanca as part of the wider Nexus AI Factory.TechAfrica NewsYates County residents pushed back against a proposal to turn a former crypto plant on Seneca Lake into an AI data center.RochesterFirstWHEC.comOpen 55 stories →
Hyperscalers · 11 storiesCapex, power procurement, model-capacity deals.Microsoft weighs Azure cost cuts as Huang repeats $3–4 trillion forecastA report said Microsoft is considering cost cuts in Azure amid rising AI data center spending. Analysts and executives offered competing reads on how long the spending lasts.Hyperscalers, indexed over the last 10 sessionsClosing prices on each company's home exchange, indexed to 100 at the first session shown, for the 5 hyperscalers that moved most over the window.Yahoo Finance reported Microsoft is weighing cost cuts in its Azure unit amid soaring AI data center spending.Yahoo FinanceJensen Huang reaffirmed his forecast that AI infrastructure spending will reach $3 trillion to $4 trillion by 2030.The Motley FoolThe Globe and MailThe Register reported that cheap Chinese and open-source LLMs could trigger a trillion-dollar reset in AI infrastructure.The RegisterOpen 11 stories →
Semiconductors · 37 storiesAccelerators, memory, foundry, packaging.Tower and Semtech expand SiGe capacity deal as memory chipmakers sell offSemtech will invest in dedicated SiGe capacity at Tower for AI data center interconnects. US chip stocks extended their slide on a China memory report and AI spending concerns.Chipmakers this window, by move0 of 8 tracked chipmakers rose over the window; the change is close to close on each company's home exchange, in its own currency.Tower Semiconductor and Semtech will invest in dedicated SiGe capacity at Tower for AI data center interconnects, with Semtech securing capacity in three countries.TradingViewStock TitanYahoo FinanceThe Nasdaq dropped as memory chipmakers sold off, and Reuters said a China memory report and AI spending jitters hit the sector.Yahoo FinanceReutersBenzinga reported an analyst says Alphabet could unlock $100 billion from AI chips by 2028.BenzingaOpen 37 stories →
Semi equipment · 16 storiesLithography, deposition, etch, test, export rules.Equipment shares fall before ASML results; Hanmi shows HBM4 toolApplied Materials, Lam Research and KLA fell on profit-taking. ASML raised its 2026 revenue outlook as its earnings tested the group's momentum.Equipment stocks' price declines in the profit-taking slideApplied Materials fell 7%, steeper than the 4% drops at Lam Research and KLA, as investors took profits after a strong year; all three figures come from a single report.Applied Materials sank 7% and Lam Research and KLA dropped 4% as investors took profits.24/7 Wall St.ASML raised its 2026 revenue outlook, with ChartMill saying its earnings put equipment stocks' momentum to the test after a pullback.GuruFocusChartMillHanmi Semiconductor unveiled HBM4 equipment at SEMICON West and is targeting the US market.BigGo FinanceOpen 16 stories →
New research
Training-free reasoning methods and MoE serving work lead a day of efficiency papers
Hybrid Reasoning Agents from Together, Stanford and UIUC gained 10.4% on GPQA, MATH and agentic benchmarks while cutting inference cost 2.7×–3.7×, with weights and code released; DoubtLess cut reasoning tokens by 31% without accuracy loss on a 32B model. BELT cut KV cache memory by 70% and FastEEP sped MoE elastic scaling by 7.1×. A reproducibility study of five factuality alignment methods found standard approaches fail to generalize.
Which fields movedpapers today ● vs 30-edition average ○
01345 papers
How open was it
Weights released3 / 8
Code released5 / 8
Who published0 lab · 1 mixed · 7 academic
Reported gains today, in percent
3 of 8 kept papers report their headline result as a percent gain; papers reporting in other units are not put on this axis.
The papers · headline result, scale, what was released
AI + Energy & Sustainability
Frontier AI power on the record, by country8 rows · 1 source
Stated power capacity summed across the frontier sites Epoch AI records in each country — where the grid load of the build-out actually sits.Epoch AI · Frontier Data Centers Hub
When the energy stories landed
2 charts built from 1 stated figure across 18 stories, the first planned by the composer from the record. Also had the data for pressure and relief by topic. Held back: power committed in deals that state a figure (no deal in this edition states the capacity it commits); power disclosed per edition · last 7 (only 0 of the last 7 editions put a number on it); other figures on the record today (only 1 other figure on the record — the card needs 3); what today covered · topic by region (today’s energy stories fall in one topic or one region — a grid needs two of each).