The Investor

The mechanics matter more than the marks here, which is the polite way of saying read the terms before you read the headline. SambaNova nearly sold itself last year, and its eleven-billion-dollar round is underwritten by an Intel partnership plus the assumption that some strategic buyer eventually pays Groq-style money. FuriosaAI turned down Meta's eight hundred million and is raising north of five hundred million at a pre above two billion, led by Korea's DSC Investment, with founders anchoring their exits to Nvidia-scale outcomes — Meta, having been told no, consoled itself with Rivos. Nuvacore is the purest pedigree trade of the bunch: ex-Apple chip engineers who sold a startup to Qualcomm, Sequoia-seeded three months ago, now raising two hundred million and change. Few of them have material revenue.

The tell is the public tape. As the privates reflated, SK Hynix ADRs fell 9.3% in a Korea semi rout and capital rotated into Apple as the capex-light AI haven. Publics are repricing AI hardware risk down while private silicon reprices up. One of them is wrong, and the privates lag.

The thesis holds where the play is to complement the GPU, not replace it. d-Matrix on inference and Ayar Labs on optical interconnect sit where strategic-buyer logic is credible for more than one acquirer. SpaceX's Terafab, plus its FTC-approved Mesh Optical acquisition, drops a new strategic buyer and a new competitor into a single entity.

The bear case is arithmetic, which is the hardest kind to argue with. The whole reflation assumes the Groq deal repeats, and Nvidia bought Groq defensively at maximal cash and maximal threat perception, which is not a market clearing price. One down-round or one failed process across this cohort in eighteen months and the momentum bid evaporates, leaving prototype-stage hardware trading at software multiples.

Treat this as a liquidity window, not a floor. If you hold, the eleven-billion and five-billion comps are a gift for partial secondaries; take them. If you are entering, require strategic-buyer optionality across at least two credible acquirers — Nvidia, Meta, Intel, Qualcomm, SpaceX — plus production milestones. Benchmark decks do not count.

New York built the mechanism that makes this contagious: an executive order blocking environmental permits for 50MW+ facilities, and a legislative bill at 20MW+ waiting on Hochul's signature. The politics run 46% for, 21% against, on utility bills and grid stress that exist in every populous state, so twelve-plus states already have comparable bills filed. When the binding constraint stops being chips (buyable) and becomes permits and power contracts (not accelerable at any price), value migrates to whoever is holding pre-permitted capacity, behind-the-meter gas, grid interconnect. One hyperscaler campus needs seven new natural-gas plants. That is the whole sentence.

Demand delivered its own verdict. IBM warned Q2 would disappoint because clients diverted mainframe budgets to AI hardware, and the stock fell 23% in a single session — its worst drop in decades — to $204B, which is below Palo Alto Networks and below CrowdStrike. The same reallocation that rewards the winners now punishes the perceived losers on the same afternoon. Anything AI-adjacent but not AI-native — legacy IT, generic enterprise resellers, consulting-flavored transformation plays — is a re-rating candidate on its next print, or rather on the first print the market reads carefully.

The memory squeeze completes the picture. DRAM and HBM reallocated to data centers drove smartphone shipments down 11% to a 13-year low, and only Samsung at 24% and Apple at 20% gained share while the mid-tier got euthanized. That is demand destruction, which is the tell that the reallocation is physical rather than narrative. AWS is constrained enough that startups are defecting to alternative clouds, which opens a time-boxed logo-grab window for neoclouds and inference specialists. Time-boxed being the operative word.

The buildout continues in aggregate — China AI exports up 27%, Intel committing €5B to an Irish fab — so the argument is not whether but where value accrues, and the sources disagree violently. The resolution is a barbell: scarce physical inputs on one end, capex-light AI monetizers on the other, capex-heavy middlemen and legacy IT squeezed in between. This is probably wrong on the timing. It is likely right on the shape.

The coordination is the new part, and the data underneath it. Nadella, Benioff, and Karp all warned inside 72 hours that Anthropic and OpenAI harvest customer knowledge to compete up the stack — Karp put it bluntly: they'll 'take the alpha of my business, transfer it into their model weights, and compete against me.' The messaging is self-serving, of course (Microsoft has no frontier model; Salesforce runs on both providers), but the anxiety it channels is a real demand signal for a neutral middleware layer. Dedicated instances are already 'the norm' at financial institutions. That entrenches the hyperscalers as the toll booth.

The margin data hardened this week. Open-weight models are 29% of Vercel Gateway token volume at under 4% of spend — a 7-to-1 volume-to-revenue gap, which is the floor collapsing under commodity inference. Meta enters the API market on price with a 1M-context model, Zuckerberg calling current models 'too expensive.' Distilling a frontier model into a deployable specialist runs about ~$250. Three labs shipped computer-use in the same week; three coding agents shipped identical /goal features inside a fortnight. Feature moats now have a half-life under six months.

Value sticks at distribution and evals, or rather, wherever the customer can't easily switch. Codex hit 6M users adding 1M/day, roughly 3x Claude Code's last-reported figure, which says the coding-agent war is won on distribution and not on benchmarks. a16z planted its flag: pure SaaS is 'nearly uninvestable,' and the next trillion accrues to transformation companies that encode incumbent processes into agents, with proprietary eval suites as the moat (coding is ~99% of AI revenue because code self-evaluates). Read the memo as a preview of their next checks. Read it also as a leading indicator of multiple compression for eval-less subscription software.

The diligence change is smaller and more useful: token-level COGS is now a required IC input. Identical workloads cost 73% more on one vendor's tokenizer than another's, and agent loops multiply consumption on top of that. Headline ARR on a thin wrapper can sit over structurally negative unit economics. Read the footnote.

The freeze comes with its own justification, and that justification is what kills the old thesis. Government research found CMMC would push contractors out of the defense industrial base, with SMB compliance costs exceeding $7B annually and roughly 100 accredited assessors on hand for 100,000-plus companies. That scarcity was sold to investors as a moat. It is now the stated reason to tear the program down. Phase 2 was meant to become mandatory on November 10, 2026. Instead a Reform Task Force reports within 60 days, and DOD officials explicitly declined to rule out full cancellation.

The rotation is legible, or rather it is legible if you read what the pause actually does. DOD reverts to NIST 800-171 Rev 2 self-assessments — a lower, cheaper bar that shifts value from third-party certification to GRC automation, evidence collection, and continuous monitoring, the tooling that makes self-attestation audit-proof. Those tools win whether CMMC returns reformed or dies quietly. What is not winning is the C3PAO assessor business and the certification-dependent readiness SaaS, which now carry an existential binary that resolves this quarter.

The offsetting signal sits in the same domain. Europe's OT security demand got a kinetic catalyst. The FSB Center 16 / Turla campaign — a 16-year operation across eight-plus nations — ended in a December attack on Poland's energy grid that left 500,000 people without heat, which triggered coordinated EU/UK sanctions. Critical-infrastructure security has become a board-level, national-security spend line. That is where GovCon-adjacent capital rotates when the certification trade goes cold.

The deal mechanics are the interesting puzzle here. Any active compliance term sheet needs regulatory-MAC protection until the Task Force reports. Any portfolio company with material CMMC-dependent forecast revenue needs a marked-down base case that assumes cancellation, not delay. The 60-day clock is a rare thing in this business: a regulatory catalyst with an underwritable date.