brianletort.ai
All issues

The AI Stack Weekly

Issue 06 · Week 22 of 2026.

/Industry brief · ~7 min read/Public sources onlyDownload brief

The Bottom Line

A quiet release week resolved last week's two biggest open questions — and moved the contested ground to optics and the agent control plane.

Flywheel arcAll three lenses

After W21's product deluge, W22 was a confirmation week: the two binary reads left open last week both resolved, and where almost nothing shipped, the leaderboard and the balance sheet still moved. On software, Claude Opus 4.8 (May 28) retook #1 on the Artificial Analysis Intelligence Index (61.4 vs GPT-5.5's 60.2) with a >10-point SWE-Bench Pro lead — at flat list pricing — while Gemini 3.5 Pro slipped to June, leaving Anthropic the unchallenged closed frontier. On capital, the W21 watch item resolved cleanly: Anthropic's round closed May 28 as a $65B Series H at a $965B post-money, the largest private AI raise on record (though ~$15B is previously-committed hyperscaler capital, not fresh cash). On hardware, the other open risk closed too — Samsung's union ratified its wage deal on May 27 (73.7% approval), formally averting the strike that threatened HBM4 packaging, so the binding constraint stays supply (CoWoS + HBM4 + surging DRAM pricing), not demand. The genuinely new front was networking: four independent vendors (GlobalFoundries, Wiwynn, Credo, Edgecore) pushed co-packaged optics and all-photonic fabric from demos toward deployable product in-window, with one analyst house putting a ~$154B TAM on optical interconnect — a textbook Gilder arc where per-GPU bandwidth, and the dollar content that carries it, scales faster than compute. So-what for the four audiences: boards should treat the HBM4 supply scare as closed and the custom-silicon counter-flywheel (ASIC shipments +44.6% YoY vs +16.1% for merchant GPUs) as the durable margin question; investors should read the Anthropic close as the answered binary and discount the louder unverified headlines (a '$45B Anthropic-pays-xAI-through-2029' filing the principal himself recut to a ~180-day lease; a secondary-sourced '$1T OpenAI September IPO'); architects should re-baseline coding-agent evals on Opus 4.8 but pin effort levels first; operators should put agent governance and the data-cloud control plane (Snowflake's Natoma/MCP buy, Salesforce's $1.2B Agentforce ARR) on this quarter's procurement roadmap.

JevonsMetcalfeGilderSoftwareJevonsHardwareHuangNetworkingMetcalfe + Gilder

The three lenses

What moved this week, and what to do about it.

12 events across the flywheel — 4 software, 4 hardware, 4 networking.

Software.

  • Anthropic released Claude Opus 4.8 (claude-opus-4-8) at flat pricing ($5/$25 per Mtok, fast mode ~3x cheaper) plus a Dynamic Workflows research preview; retakes the Artificial Analysis Intelligence Index lead at 61.4 (vs GPT-5.5 60.2) and SWE-Bench Pro 69.2%, though it loses Terminal-Bench 2.1 to GPT-5.5

    anthropic.com/news/claude-opus-4-8, artificialanalysis.ai, TechCrunch

  • Liquid AI shipped open-weight LFM2.5-8B-A1B, an on-device MoE (8.3B total / 1.5B active, 128K context) decoding ~253 tok/s on a laptop with day-one llama.cpp / MLX / vLLM / SGLang support

    liquid.ai/blog/lfm2-5-8b-a1b, MarkTechPost

  • StepFun released open-weight Step 3.7 Flash (198B sparse MoE, ~11B active, 256K context) under Apache 2.0 at ~400 tok/s, claiming SWE-PRO 56.3 and 98%+ on tau-squared-bench, positioned for agentic / coding / search workloads

    StepFun HuggingFace model card, Digg

  • A Financial Times report showed the open-source 'Heretic' abliteration tool (~17.8K stars) stripping safety guardrails from open-weight models such as Llama 3.3 and Gemma 4 within minutes, spotlighting the open-weight safety floor ahead of the EU AI Office's Aug 2 Article 92 evaluation powers

    Financial Times (via Singularity.Kiwi), Lawfare

What this means

Two arcs run in parallel: (1) a software->hardware Jevons signal — the closed frontier pushed the ceiling (Opus 4.8) while a same-day open-weight efficiency wave (Liquid on-device 128K at 253 tok/s, StepFun 400 tok/s under Apache 2.0) pushed cheap-and-good inference outward, so architects should re-baseline cost-per-task now rather than wait for price cuts; (2) capability-risk procurement hardens as trivially abliterated open weights lower the deployed safety floor just as the EU AI Office's Article 92 access powers arrive Aug 2 — boards should treat open-weight governance as a vendor-risk dimension, not a compliance afterthought.

Hardware.

  • Samsung's union ratified its 2026 wage-and-bonus deal with 73.7% approval on 95.5% turnout, formally averting the strike that had threatened the testing/packaging lines feeding HBM4 supply for next-gen AI accelerators

    Bloomberg, Korea Herald, union joint-bargaining-team announcement

  • TrendForce's in-window DRAM Market Bulletin reported seller-driven contract price hikes, said HBM prices will surge once a temporary margin lag clears, and warned fab-expansion lead times will keep memory supply bottlenecked 'for years,' with HBM4 the gating 2026 constraint

    TrendForce DRAM Market Bulletin (RP260527ZM)

  • Alchip's CEO said custom-ASIC revenue growth will outpace the GPU market, aligning with TrendForce data showing cloud custom-silicon shipments tracking +44.6% in 2026 vs +16.1% for merchant GPUs — the first AI-era year custom chips meaningfully outgrow GPUs, with ASICs ~27.8% of AI-server shipments

    Alchip CEO remarks + TrendForce, via TechTimes

  • AMD began the production ramp of its 6th-gen EPYC 'Venice' (Zen 6, up to 256 cores, 1.6 TB/s memory bandwidth) — the industry's first HPC product on TSMC's 2nm node and built on SoIC-X / CoWoS-L packaging, feeding the MI450X-based Helios rack into the Computex run-up

    AMD IR press release, Tom's Hardware, Wccftech

What this means

A constraint-and-cadence week, not a launch week: the binding limits on the hardware flywheel are packaging (CoWoS) and memory (HBM4 plus surging DRAM/HBM pricing), and the Samsung labor resolution defused the single biggest near-term HBM4 supply shock just before the June 1 GTC Taipei keynote. Underneath, the custom-silicon counter-flywheel is real (ASIC shipments +44.6% vs +16.1% GPUs) and AMD's 2nm/SoIC-X Venice ramp shows rivals fighting for the same TSMC node-and-packaging stack that gates every accelerator from wafer to rack — investors with concentrated merchant-GPU exposure should diversify into ASIC co-design and advanced packaging.

Networking.

  • GlobalFoundries launched its SCALE co-packaged silicon-photonics module, claiming the first platform to comply with and exceed the Optical Compute Interconnect (OCI) MSA, with 50G/100G micro-ring modulators and demonstrated 8-lambda/16-lambda bidirectional DWDM for optical scale-up interconnect

    GlobalFoundries release via ET Datacenters

  • Wiwynn unveiled a CPO-based optical scale-up rack design (Ayar Labs TeraPHY engines, ELSFP SuperNova remote light sources, GUC 2.5D packaging) ahead of Computex 2026, framing co-packaged optics as the path to move AI scale-up fabric beyond copper limits

    Wiwynn press release

  • Credo closed its acquisition of DustPhotonics, adding silicon-photonics PIC technology to build a vertically integrated interconnect stack (SerDes + DSP + SiPho + system integration) spanning 800G, 1.6T and 3.2T near-packaged and co-packaged optics for scale-out and scale-up AI networks

    Credo via Business Wire / Morningstar

  • Edgecore (Accton) and NTT announced 'Edgecore Open Fabric: Built for IOWN,' an all-photonics fabric anchored by an IOWN DCI Gateway (LCoS optical switch doing DWDM/ROADM wavelength routing with zero O-E-O conversion) over a 102.4 Tbps Broadcom-silicon, SONiC-compatible spine for standards-based multi-site optical interconnect

    Edgecore/Accton via Business Wire & Telecompaper

What this means

W22 was an optical-interconnect week: four independent vendors pushed CPO and all-photonics from component demos toward deployable fabric, and one analyst house put a ~$154B TAM on it — a textbook Gilder arc where per-GPU bandwidth (and the optical dollar content that carries it) scales faster than compute itself, so value capture migrates into the fabric/optics layer that gates whether GPUs can act as one machine. As OCI-MSA and IOWN-APN standardization let more nodes interconnect coherently across racks and sites, the Metcalfe payoff becomes the explicit design target — networking architects should treat optical scale-up, not just scale-out, as a near-term procurement line.

Capital flow

Money in, revenue out.

4 categories tracked. Capital deployment up in 3 of 4; revenue follows at multiples of 0.21 to 0.6.

The four-category scorecard. Where capital is going in, where revenue is coming out, and how much of it is real. The one chart for the boardroom.

  • Frontier Labs

    OpenAI, Anthropic, Google DeepMind, xAI

    Capital In

    ~$90B ↑

    vs ~$52B

    Revenue Out

    ~$20B ↑

    vs ~$48B

    Burn / Rev

    ~1.3x

    Movement

    Anthropic closed a $65B Series H on May 28 at a $965B post-money — the largest private AI raise on record

  • Hyperscaler-Hosted

    Azure-OpenAI, AWS-Anthropic, Google Cloud-Gemini, Oracle-OCI

    Capital In

    ~$180B ↑

    vs ~$60B

    Revenue Out

    ~$60B ↑

    vs ~$20B

    Burn / Rev

    ~0.3x

    Movement

    Anthropic's round confirmed ~$15B of previously-committed hyperscaler capital, deepening lab-cloud entanglement

  • Neoclouds

    CoreWeave, Nscale, Crusoe, Lambda, Fluidstack, IREN

    Capital In

    ~$12B ↓

    vs ~$30B

    Revenue Out

    ~$5B ↑

    vs ~$3.0B

    Burn / Rev

    ~3x

    Movement

    IREN signed a ~$9.7B multi-year Microsoft AI-cloud deal (May 29) plus a ~$5.8B Dell GPU purchase

  • On-Prem / Hybrid

    Enterprise GPU clusters, sovereign and national programs, Cisco / Dell / HPE

    Capital In

    ~$90B ↑

    vs ~$45B

    Revenue Out

    ~$35B ↑

    vs Indirect

    Burn / Rev

    ~2x

    Movement

    SoftBank committed up to EUR 75B to build 5GW of AI data centers in France (May 30, Choose France summit)

Burn-to-Revenue is revenue divided by committed capital. Lower means more capital is going out than coming in.

See the full breakdown

Signal vs noise

What’s real, what’s noise.

5 claims this week — 3 signal, 2 noise.

Each claim is scored 1–5 on source quality and triangulation. Anything 2 or below is flagged as noise. Where consensus is wrong, we say so.

  • 4 / 5

    Anthropic closed a $65B Series H on May 28 at a $965B post-money valuation, becoming the world's most valuable private AI company.

    Sources: Company press release (May 28) + TechCrunch + Bloomberg (which the prior week flagged a '>$30B / >$900B' round closing 'as soon as next week' — it closed larger). Caveat: private round, not SEC-audited; ~$15B is previously-committed hyperscaler capital, so fresh cash is materially below the $65B headline.

    Real and historic — the largest private AI raise on record and a clean answer to the W21 watch item (yes, it closed in-window). It resets the private-market ceiling ahead of an IPO race and locks strategic silicon/cloud partners onto the cap table. What would change the read: an eventual S-1 separating fresh cash from in-kind compute commitments.

  • 2 / 5 — noise

    Anthropic has committed ~$45B to xAI/SpaceX for compute 'through 2029.'

    Sources: SpaceX S-1 (filed May 20, SEC) states $1.25B/month through May 2029 (~$45B cumulative); but the principal publicly recast it on May 28 as a '180-day lease with 90-day mutual cancellation,' implying ~$7.5B. CNBC/IBD flagged the filing-vs-statement divergence.

    NOISE / overstated. The 'through-2029 / ~$45B' figure is a filing ceiling, not a binding commitment — either-side 90-day termination means only the monthly run-rate is reliable, and the principal himself walked back the duration. It distorts perceived cross-lab compute dependence. What would change the read: an amended S-1 or counterparty confirmation of a minimum term.

  • 2 / 5 — noise

    OpenAI filed to IPO at ~$1T with a September 2026 listing.

    Sources: Reuters/Axios/Fortune (confidential draft S-1 reported ~May 22). Caveat: a confidential DRS is not visible on EDGAR; the valuation and timing are reporter estimates, and the CFO had previously signaled a 2027 preference.

    Mostly hype on specifics. The confidential filing is a real signaling step, but '$1T' and 'September' are secondary-sourced extrapolations that conflict with prior company guidance. It matters as the anchor for the entire AI-IPO calendar. What would change the read: a public S-1 on EDGAR.

  • 3 / 5

    A new benchmark (DeepSWE) shows SWE-Bench Pro is contaminated, reordering the coding leaderboard.

    Sources: VentureBeat + Datacurve (released May 25; dataset, harness, and trajectories published on GitHub). Caveat: single primary source with commercial interest in a competing benchmark; verdicts are LLM-judged on modest samples.

    Credibly real as a methodological finding — the .git-history exploit (cited at 18-25% of certain passes) is reproducible from the published artifacts — but self-interested and not yet independently replicated. It implies the open-vs-closed coding gap is partly an artifact. What would change the read: third-party replication of the contamination rate.

  • 4 / 5

    SoftBank committed up to EUR 75B to build 5GW of AI data-center capacity in France (Choose France, May 30).

    Sources: SoftBank press release (May 30) + summit coverage. Caveat: phased and back-loaded (phase 1 ~EUR 45B / 3.1GW by 2031); an announced commitment, not yet financed or under construction.

    Real and the largest single European AI-infrastructure pledge to date, validating the sovereign-AI capital thesis — but the capacity is years out and privately funded rather than state capex, so near-term grid/revenue impact is limited. What would change the read: financing close, groundbreaking, or grid-interconnection milestones.

Early warning panel

The levers we monitor.

10 metrics tracked — 2 rising, 0 falling, 8 steady.

Current vs prior period. Each metric has a threshold where the read materially changes — this panel flags the inflection before it lands in headlines. Click any metric for the methodology and this-week read.

  • Frontier lab cash position (avg months runway, top 3)

    ~33-36 mo↑vs ~30 mo

    Threshold: <18 mo triggers re-rating risk

    What this measures

    Top 3 frontier labs (OpenAI, Anthropic, Google DeepMind) by disclosed runway. Anthropic's $65B Series H closed in-window (May 28, $965B post-money), materially extending the top-3 average on top of the leader's prior cumulative committed capital. Boards should not assume frontier-lab funding pressure as a forcing function for short-term commercial concessions — the runway just got longer.

  • Hyperscaler capex / AI revenue ratio (top 4 weighted)

    ~5.0-5.2→vs ~5.0-5.2

    Threshold: >6.0 invites investor pushback at next earnings

    What this measures

    Top 4 hyperscalers (MSFT, GOOG, META, AMZN) weighted aggregate of total capex divided by AI-attributable revenue. No within-window prints — all top-4 readings came at late-April earnings (~$725B 2026 capex guide), so this is carried flat. Investors monitoring a 'capex bubble' should keep the hypothesis on power / HBM4 supply constraints, not demand.

  • CoreWeave revenue backlog

    $99.4B→vs $99.4B

    Threshold: Conversion velocity matters more than gross figure

    What this measures

    Booked but unrecognized revenue. The $99.4B audited figure (as of Mar 31, reported May 7) is unchanged; next print is Q2 in early August. Operators evaluating neocloud counterparty risk should keep watching conversion velocity over the headline backlog number.

  • NVIDIA Q-over-Q data center revenue

    $75.2B (Q1 FY27)→vs $75.2B (Q1 FY27)

    Threshold: Q2 FY27 guide $91B implies further +21% QoQ

    What this measures

    Q1 FY27 Data Center revenue of $75.2B (+21% QoQ, +92% YoY) was reported May 20 (prior window); Q2 guide is $91B with zero China DC compute assumed. No within-window change. HBM4 supply — with the Samsung labor risk now removed (May 27 ratification) — remains the binding constraint, not demand.

  • Open vs closed gap on SWE-Bench Pro (coding)

    Closed +~19pp (audit caveat)→vs Closed +19pp

    Threshold: Sustained open lead reshapes enterprise procurement

    What this measures

    Top closed (gated Claude Mythos Preview 77.8%) vs top open (~58.6%) is roughly unchanged on the May 27 board. But a May 25 third-party audit (DeepSWE/Datacurve) found Claude Opus models exploited a .git loophole in 18-25% of certain passes — the real open-vs-closed gap may be overstated. Architects should treat single-benchmark superiority claims with more skepticism and pilot open self-host options before signing multi-year closed contracts.

  • Sovereign AI commitments (count / aggregate $)

    ~13 / ~$160B+↑vs 12 / ~$80B+

    What this measures

    SoftBank's up-to-EUR 75B / 5GW France pledge (May 30, Choose France) was added in-window, roughly doubling the curated aggregate. Counts are analyst-curated rather than a single audited figure. Operators with EMEA workloads should treat European sovereign compute as an increasingly credible landing zone, while pricing in multi-year build timelines.

  • PJM 2026/27 capacity auction price ($/MW-day)

    $329.17→vs $329.17

    Threshold: 11x in 24 months — power is the new binding constraint

    What this measures

    The 2026/27 BRA cleared at the FERC cap ($329.17, July 2025) and takes effect June 1, 2026; no new auction in-window. Architects should not assume near-term price relief from forward auctions; budget capacity at-cap through 2028.

  • Time-to-power, busiest US markets (months)

    60-84 (new PJM); 36-48 (existing PJM queue)→vs 60-84 (new PJM); 36-48 (existing PJM queue)

    What this measures

    Months from new-load interconnection request to energization. PJM data confirms ~7-year new-build timelines, essentially flat in-window, but the bottleneck has shifted downstream: substation transformer lead times ticked up from ~150 to >160 weeks in 2026. Architects should pre-commit power — and now long-lead grid equipment — before pre-committing GPU SKUs.

  • Cost-per-task, frontier reasoning model

    ~$0.10-$0.15 (effective)→vs ~$0.10-$0.15 (effective, with hidden reasoning tokens)

    Opus 4.8 fast mode dropped ~3x but no verifiable per-task reading in-window

    What this measures

    Median cost across frontier-tier reasoning models for a benchmark complex task. No verifiable within-window reading, so carried from W21 (flagged low-confidence). Opus 4.8's fast mode dropped ~3x in list terms; operators running agents at scale should re-benchmark on cost-per-task, not list price, once independent figures land.

  • Custom silicon share of incremental AI compute

    ~33-36%→vs ~33-36%

    Threshold: >35% materially compresses merchant GPU pricing

    What this measures

    No new primary reading in-window, but consistent secondary data (TrendForce/SemiAnalysis) shows ASIC AI-server shipments ~27.8% of the 2026 market growing +44.6% YoY vs +16.1% for merchant GPUs. Investors with concentrated NVIDIA exposure should diversify into ASIC co-design (Broadcom, Marvell) and advanced packaging / power.

Predictions

What we expect next.

5 predictions for the next 30-90 days, confidence 60%-75%.

Each prediction is falsifiable, time-bounded, and tied to a specific signal we will watch. Future issues score these hit, miss, partial, or pending and build a public track record.

Prediction 01

65%

confidence

Capital

No frontier lab (Anthropic or OpenAI) files a publicly visible S-1 on SEC EDGAR before August 31, 2026, keeping the IPO race at the confidential-DRS stage.

Deadline: By August 31, 2026

Trigger: SEC EDGAR public filings; confirmed public S-1 vs confidential DRS reporting from Reuters / Bloomberg / The Information.

Prediction 02

60%

confidence

Software

Gemini 3.5 Pro reaches general availability by June 30, 2026 and scores AA Intelligence Index >= 61, contesting Claude Opus 4.8's fresh lead.

Deadline: By June 30, 2026

Trigger: Google / DeepMind GA announcement; Artificial Analysis leaderboard update.

Prediction 03

75%

confidence

Hardware

At GTC Taipei / Computex (June 1), NVIDIA reaffirms Vera Rubin production starting in 2H 2026 and frames HBM4 + CoWoS as the binding supply constraint rather than demand.

Deadline: By June 7, 2026

Trigger: NVIDIA GTC Taipei keynote; press coverage; investor notes.

Prediction 04

65%

confidence

Networking

At least two of (Credo, Marvell, Broadcom) cite co-packaged-optics or 1.6T design wins in their next quarterly earnings, validating the W22 optical-fabric push.

Deadline: By August 31, 2026

Trigger: Q2 earnings calls and investor decks from optical/interconnect vendors.

Prediction 05

60%

confidence

Power

A major hyperscaler or sovereign program announces a new behind-the-meter or >1GW power-procurement deal (SMR, gas, or grid) by August 31, 2026, as time-to-power stays the binding US constraint.

Deadline: By August 31, 2026

Trigger: Utility / PPA announcements; hyperscaler energy disclosures; sovereign program financing milestones.

Track record

Scoring prior predictions.

5 prior predictions: 0 hit, 0 miss, 1 partial, 4 pending. Hit rate 0%.

5 predictions across issues so far. Hit rate: 0%. Hits 0, misses 0, partials 1, pending 4.

Prediction 01

70%

confidence

Hardware

NVIDIA Q2 FY27 (August 2026) prints Data Center revenue at $90B-plus, confirming the $91B Q2 guide is conservative and supply-constraint thesis holds.

Deadline: By August 31, 2026

Trigger: NVIDIA Q2 FY27 earnings release; HBM4 supply commentary post-Samsung ratification vote; hyperscaler Q2 capex prints.

pending— Deadline not reached; Q2 FY27 prints in early August. The Samsung HBM4 supply risk closed favorably (May 27 ratification), supporting the supply-not-demand thesis going into GTC Taipei (June 1).

Prediction 02

80%

confidence

Capital

Anthropic round closes at a final post-money valuation between $900B and $1.1T with at least three of (Sequoia, Dragoneer, Altimeter, Greenoaks) named as lead investors publicly.

Deadline: By June 13, 2026

Trigger: Anthropic primary disclosure; SEC Form D if applicable; lead-investor naming via Bloomberg / Reuters / The Information.

partial— Valuation condition HIT: the round closed May 28 as a $65B Series H at a $965B post-money, inside the $900B-$1.1T band. The named-lead-investor condition is not yet publicly confirmed in-window, so scored partial pending the full lead roster before the June 13 deadline.

Prediction 03

60%

confidence

Software

At least one Fortune 500 enterprise discloses standardization on a single agent-platform runtime (Claude Code / Codex / Antigravity 2.0 / Grok Build) for greater than $50M annualized commitment.

Deadline: By August 31, 2026

Trigger: Enterprise architecture announcements; vendor earnings color; F500 case studies.

pending— No qualifying >$50M single-runtime disclosure yet. Momentum is supportive — Claude Code shipped Dynamic Workflows and Cognition raised at $26B — but no named F500 standardization at the threshold has been disclosed in-window.

Prediction 04

70%

confidence

Networking

At least three of (Marvell, Broadcom, NVIDIA networking, Arista) post Q2 networking-segment revenue growth greater than 40% YoY, validating fabric momentum into Q2 prints.

Deadline: By August 31, 2026

Trigger: Q2 earnings prints from networking-focused vendors; NVIDIA Q1 already at +199% YoY sets the bar.

pending— Deadline not reached; Q2 prints pending. The in-window optical-fabric wave (GlobalFoundries SCALE, Wiwynn CPO, Credo/DustPhotonics close, Edgecore IOWN) is directionally supportive of the thesis.

Prediction 05

60%

confidence

Software

Gemini 3.5 Pro launches by July 15, 2026, scoring AA Intelligence Index greater than or equal to 60, taking the closed-frontier intelligence lead from GPT-5.5.

Deadline: By July 15, 2026

Trigger: Google blog / DeepMind announcement; Artificial Analysis leaderboard update.

pending— Did NOT ship in-window (W22); Google reaffirmed a June GA window. Still inside the July 15 deadline, but it now must also clear Opus 4.8's fresh 61.4 lead, not just GPT-5.5's 60.2.

Watchlist

On the radar this week.

5 catalysts to watch, starting Jun 1.

Specific catalysts that would change the read materially. Watching these tells us whether the thesis is strengthening or weakening.

  • Jun 1

    NVIDIA GTC Taipei / Computex keynote

    Vera Rubin and Feynman cadence plus HBM4/CoWoS supply color will set the hardware-flywheel baseline for 2H 2026.

  • Jun 2026

    Gemini 3.5 Pro general availability

    Its arrival would directly contest Opus 4.8's fresh AA Index lead and reset the closed-frontier scorecard.

  • Jun 1

    PJM 2026/27 capacity prices take effect

    The $329.17/MW-day cap-bound rate begins billing June 1, hard-coding power as the binding cost constraint for the year.

  • Jun 2026

    Frontier-lab S-1 / IPO signaling

    Whether any public S-1 appears on EDGAR will separate the real IPO calendar from the secondary-sourced '$1T September' headline.

  • Jun 2026

    Computex optical / CPO follow-through

    Post-show, watch whether the W22 co-packaged-optics demos convert into named design wins and shipping fabric products.

Companion reads

The rest of the spine.

The AI Stack Weekly is the cross-stack flywheel read. Pair it with the model-and-tree spine and the working framework to get the full picture.

Edits this issue

  • — Scored W21's prediction set: p23 (Anthropic round) marked partial after the round closed May 28 at a $965B post-money; the other four remain pending pre-deadline.
  • — Added Claude Opus 4.8 to the LLM Evolutionary Tree; the Model Pulse treeDelta references it.
  • — No revisions to the thesis.

About this brief

Compiled from public announcements, SEC filings, earnings transcripts, and official lab and vendor publications. Every quantitative claim is graded 1–5 on source quality. Claims graded 2 or below are flagged as noise. The thesis the brief defends is published separately and updated only when a hypothesis materially changes.

Authorship

Written by Brian Letort. Independent analysis. All sources cited are public. Not investment guidance.

Operate. Publish. Teach.