AI in Business: Anthropic at $900B as Meta funds AI by firing 8,000  May 26th 2026

Watch & Listen First

AI + a16z: Clément Delangue on the open-source AI economy · May 22 · The Hugging Face CEO argues a bubble is forming specifically around API-priced frontier LLMs, while open weights quietly absorb enterprise workloads.

SAP Sapphire 2026 Keynote — Powering the Autonomous Enterprise · May 21 · Christian Klein's keynote pitches Joule + 200 agents as the operating system for finance, HR, and supply chain — the most aggressive bet by a legacy ERP vendor yet.

Key Takeaways

• Anthropic's $900B mark resets the AI cap table. With a $50B run-rate by June, it's the first lab to credibly threaten OpenAI's pricing power and customer base.
• Nvidia's $75B data-center quarter is the floor, not the ceiling. Hyperscaler + AI-cloud spend is still doubling YoY — capex deceleration is not visible in the print.
• Meta's playbook is now the template. Cut 10% of staff, redeploy 7,000 into AI units, fund $125–145B in 2026 capex. Expect copy-paste at every Mag 7 peer.
• The agent-platform war is now a land grab. SAP, Sierra, and Anthropic are racing to be the "system of record" for agentic workflows — the moat is integrations, not models.
• Forrester's deferral signal is the early crack. A quarter of planned 2026 AI budgets is sliding to 2027 — boards want ROI, not pilots.

The Big Story

Anthropic to close $30B round at $900B valuation, surpassing OpenAI · May 22 · Bloomberg
→ Sequoia, Dragoneer, Altimeter and Greenoaks are each writing ~$2B checks, vaulting Anthropic past OpenAI's $852B in private pricing weeks before both file to go public. This is the second $30B round Anthropic has closed inside 2026 — annualized revenue went from $9B in December to a stated $50B target by end of June — meaning investors are paying ~18x forward ARR for the lab now seen as the enterprise-default Claude vendor. The strategic read: Anthropic is positioning to outbid OpenAI on compute, talent, and acquisitions before the IPO window opens in Q4, and the cap-table signal alone will accelerate procurement decisions inside every Fortune 500 AI committee that was sitting on the fence.

Also This Week

Nvidia reports record $81.6B quarter, data center revenue up 92% YoY · May 20 · CNBC
→ Data center alone hit $75.2B (Hyperscale $37.9B, AI Clouds/Enterprise $37.4B) — yet the stock slid, telling you the buy-side now grades Jensen on the second derivative.

Meta to lay off 8,000 while reassigning 7,000 into new AI units · May 20 · NPR / The Register
→ With 2026 capex guided to $125–145B (over 2x last year), Zuckerberg is openly using payroll savings and forced internal transfers to underwrite the GPU bill — the new "Year of Efficiency."

SAP unveils Autonomous Enterprise with 200+ agents at Sapphire · May 21 · SAP
→ The Joule-powered Autonomous Close Assistant promises to compress month-end from weeks to days — if it works, it's the first credible existential threat to mid-market FP&A SaaS like Workday Adaptive and BlackLine.

EVTV signs definitive merger with AZIO AI for energy-backed data center platform · May 20 · Intellizence
→ The "neocloud + behind-the-meter power" thesis is now a standalone M&A category — watch this template repeat as hyperscaler queues stretch to 2028.

Anthropic, Mistral, DeepMind and Meta each acqui-hired an AI startup in the same week · May 18–24 · StartupHub.ai
→ The Contextual AI team (20+ researchers, including RAG co-inventor Douwe Kiela) heading to DeepMind tells you the labs are paying any price to lock down agent-infra talent before the IPO bell.

From the Lab

Forrester: enterprises defer 25% of 2026 AI spend to 2027 · The Register / Stock Titan
→ Only 15% of AI decision-makers reported positive profit impact in the last 12 months. The mechanism to watch: CFOs are now the gatekeepers for AI procurement, which compresses POC cycles, kills vanity pilots, and rewards vendors who can attach to a P&L line — Sierra's $150M ARR in 18 months is the template.

The Reliability Gap: enterprise agent benchmarks · Simmering
→ Agents that hit 60% single-run accuracy collapse to 25% across eight runs in production, with a 37% lab-to-prod gap. Translation for buyers: budget for an evaluation harness, not just a license — the unit economics of "agentic" only work if you can cap rework.

Worth Reading

• BCG rode AI to $3.6B in revenue — here's how — 25% of BCG's $14.4B 2025 revenue came from AI work; a working blueprint for any services firm staring down billable-hour erosion.
• Sierra raises $950M as the race to own enterprise AI gets serious — Bret Taylor's playbook for 40% of the Fortune 50 in 18 months reads like a thesis for what an "AI-first system of record" should look like.
• Roche acquires PathAI for $750M upfront (+$300M earnouts) — Pharma is now buying its AI stack; expect a wave of regulated-industry incumbents to follow the same upfront-plus-milestones acquisition structure.

The vendors are running a different economy than their customers — and the gap is now wide enough that both sides need to start pricing it in.