Tech Digest – May 29, 2026
Capital & Compute at Scale
Anthropic Raises $65 Billion at Nearly $1 Trillion Valuation — Financial Engineering Now Bankrolls Silicon
Anthropic raised $65 billion in Series H funding, valuing the company at $965 billion post-money — surpassing OpenAI to become the most valuable AI company. The round drew Altimeter, Sequoia, Greenoaks, and Dragoneer as co-leads, with chipmakers Samsung, SK Hynix, and Micron joining as strategic investors. This may be Anthropic’s last private round before a public listing.
Behind the equity, Apollo and Blackstone are assembling a $36 billion debt deal to buy Google TPUs for Anthropic to lease, with Broadcom backstopping the largest tranches. Even competitors face the same capital gravity: Groq is raising up to $650 million for a “second act” after a $20 billion Nvidia licensing deal cost it most of its senior engineering team.
The demand shows in the receipts. Dell reported AI server revenue of $16.1 billion for the quarter, up 757% year over year, with a $51.3 billion backlog and full-year AI revenue guidance raised to $60 billion. Dell shares surged 33% — the stock’s best day since returning to public markets.
Note: A $36 billion debt deal to buy chips and lease them back is the kind of structure normally reserved for power grids and motorways. When financial institutions treat compute like physical infrastructure, every institution bidding for digital capacity is now competing in that market.
Sources: New York Times, Anthropic, Bloomberg, Axios, SiliconAngle, CNBC
AI Capabilities
Opus 4.8 Sets New Benchmarks — Claude Code Adds Agent Swarms for Codebase-Scale Migrations
Anthropic released Claude Opus 4.8, posting state-of-the-art results: 69.2% on SWE-Bench Pro, 57.9% on Humanity’s Last Exam with tools, and 1890 on GDPval-AA. The model pairs gains in honesty with misalignment rates approaching the unreleased Mythos Preview, which Anthropic promises to bring to all customers in the coming weeks.
Alongside the model, Claude Code introduced dynamic workflows that spin up swarms of parallel subagents to execute codebase-scale migrations across hundreds of thousands of lines, using the existing test suite as the only quality gate. The combination of a stronger model and deployment-ready agent architecture moves AI-assisted development from co-pilot to migration team.
Note: A model that leads on both safety metrics and capability benchmarks simplifies vendor evaluation: the option that satisfies the compliance team is now also the strongest. That rarely happens — and it won’t go unnoticed by procurement offices weighing AI contracts.
Sources: Anthropic, Claude Code Blog
European Sovereignty
EU Drafts Emergency Powers to Override Chip Contracts During Shortages
A revised EU Chips Act draft, due June 3, would give the European Commission emergency powers to intervene in semiconductor supply chains during shortages. Under a declared “crisis stage,” the Commission could force chipmakers to prioritise crisis-critical orders — overriding existing private contracts — and fine companies up to €300,000 for withholding supply-chain information. The draft also enables common purchasing across member states and restricts US cloud providers from holding sensitive government data.
Note: Contract law is the foundation of procurement planning. When the Commission can override a chip contract during a crisis, every institution’s supply agreements acquire an asterisk — one that could work for or against them, depending on how “crisis-critical” their order ranks.
Sources: Financial Times, Euronews
Mistral Chases Superintelligence — CEO Warns Europe Has a Two-Year Window
Mistral AI is accelerating its push toward superintelligence, with CEO Arthur Mensch warning that Europe has roughly two years to establish independent AI infrastructure before dependency on American tech becomes permanent. The French startup is investing $4.7 billion in data centres in France and Sweden, including an €830 million facility south of Paris equipped with 13,800 Nvidia GB300 chips. Mistral has secured government contracts across Europe, North Africa, and the Middle East.
Note: Two years. That’s one procurement cycle for most public institutions — and Mensch’s estimate of how long Europe has to build sovereign AI capacity before switching costs make the choice for them.
Sources: Wall Street Journal, Yahoo Finance
The Economics of Intelligence
Meta Charges for AI at $7.99/Month — Kirkland & Ellis Spends $500 Million to Build Its Own
Meta will charge for AI features for the first time under a “Meta One” subscription brand, starting at $7.99/month for deeper reasoning and more compute-heavy queries, with a Premium tier at $19.99. Testing begins next month in Singapore, Guatemala, and Bolivia, with Meta AI remaining free for casual use. The move follows OpenAI, Anthropic, and Google in treating AI compute as a metered utility.
At the other end of the market, Kirkland & Ellis — the world’s largest law firm by revenue — has set aside $500 million over four years to build a proprietary AI platform. A team of 180 tech professionals, guided by 250 lawyers, is developing the system, and outside vendors are contractually barred from selling the technology to competing firms. Chair Jon Ballis says off-the-shelf AI “isn’t high enough” for clients that expect differentiation.
Note: The AI market is splitting into two tiers: commodity (rent from Meta at $8/month) and proprietary (build for $500 million). The gap between what general-purpose AI can do and what a sophisticated buyer demands is wide enough to justify a half-billion-dollar investment. Kirkland is betting that gap won’t close.
Sources: CNBC, Financial Times, Law.com
Digital Security & Supply Chain
IBM and Red Hat Commit $5 Billion to Secure the Open-Source Supply Chain
IBM and Red Hat launched Project Lightwell, committing $5 billion and over 20,000 engineers to secure the open-source software supply chain. The initiative deploys frontier AI to identify, validate, and test vulnerability fixes at enterprise scale, delivered through commercial subscriptions. Early adopters include Bank of America, Goldman Sachs, JPMorgan Chase, Mastercard, and Visa — institutions that rely on the same open-source foundations as public infrastructure worldwide.
Note: More than 90% of Fortune 500 companies depend on open-source software. When the financial sector’s largest institutions pre-subscribe to a vulnerability clearinghouse, they’re signalling that maintaining open-source dependencies without dedicated security infrastructure is becoming untenable.
Sources: IBM Newsroom, Red Hat, Wall Street Journal
New Browser Attack Fingerprints Your Activity Through SSD Timing — No Permissions Required
Researchers at Graz University of Technology published FROST, a browser-based side-channel attack that identifies which websites and applications a user has open by measuring SSD read/write timing through JavaScript. The attack requires no permissions and no user interaction — visiting a page running the attack code is enough. It identified websites with 89% accuracy and running applications with 96% accuracy in testing. Chromium does not classify fingerprinting as a security vulnerability; Apple and Mozilla have acknowledged the issue but shipped no defences.
Note: The browser security model assumes meaningful actions require permissions. This attack needs none. Until browser vendors reclassify fingerprinting, the gap between what’s technically possible and what’s actually blocked keeps widening.
Sources: Ars Technica, Help Net Security
Pentagon Confirms Military Personnel Targeted Through Commercial Location Data
The Pentagon confirmed that US military personnel are being targeted using commercially available location data — the kind sold by data brokers and embedded in everyday mobile apps. The acknowledgement comes amid growing concern about how adversaries track troop movements, identify personnel at sensitive installations, and build intelligence profiles using data streams that are legal to purchase.
Note: If commercial data brokers can locate military personnel, they can locate anyone. Every institution permitting personal devices on premises is generating a location dataset that is for sale.
Sources: Reuters
Workforce Disruption
Wix Cuts 1,000 Jobs Citing AI — Over 134,000 Tech Roles Eliminated in 2026
Wix is cutting roughly 20% of its workforce — over 1,000 positions — citing AI’s “fast evolution” and a strong Israeli shekel that makes its workforce more expensive in dollar terms. Shares had already fallen 27% after first-quarter earnings missed expectations. The cuts place Wix alongside a wave of restructurings that have eliminated more than 134,000 tech jobs across 212 layoff events in 2026.
Note: The dual pressure — AI reducing headcount needs while currency shifts raise per-head costs — is not unique to Israel. Any institution sourcing technical services internationally faces the same squeeze from both directions simultaneously.
The common thread: infrastructure at every layer. Financial infrastructure converting capital into compute at sovereign scale. Regulatory infrastructure acquiring emergency powers over chip supply chains. Security infrastructure mobilising billions to protect the open-source code everything runs on. The entities building this move on quarterly timescales. The institutions that depend on it mostly plan annually.
