Digital Colliers Daily Briefing — June 2, 2026
Three converging developments on Monday redrew the contours of the AI industry's capital, silicon, and distribution layers. Anthropic became the first frontier lab to formally file for an IPO, doing so at a valuation that would make it the largest tech debut on record. Nvidia opened Computex by pushing into the Windows PC CPU market with an Arm-based "superchip" anchored by a new Microsoft Surface flagship. And OpenAI paired a Michigan groundbreaking for its second Stargate campus with the general availability of its frontier models on AWS Bedrock — a distribution move that lands its software inside the procurement pipelines of millions of enterprises.
1. Anthropic beats OpenAI to the IPO line at a $965B valuation

What happened. Anthropic confidentially submitted a draft S-1 to the SEC on Monday, formally beginning its path to public markets. The two-paragraph blog post did not disclose share count or pricing, citing Rule 135. The filing follows by less than a week the close of a $65 billion Series H — led by Altimeter, Dragoneer, Greenoaks, Sequoia, Capital Group, Coatue, and D1 — that lifted Anthropic's post-money valuation to $965 billion, eclipsing OpenAI's $852 billion mark from its March round. TechCrunch reports Anthropic's revenue run-rate has reached $47 billion, up from $9 billion at the end of 2025.
Why it matters. Anthropic is the first frontier lab to commit, even confidentially, to public market scrutiny. As The Register notes, the eventual S-1 disclosure will be the first audited look at the unit economics of a top-tier AI lab — a moment with the potential to either harden or puncture current AI valuations. The filing also reframes the OpenAI–Anthropic rivalry as a capital-markets race, now running parallel to SpaceX's planned June 12 debut at a reported $1.75 trillion target.
Who is affected. Amazon, an early strategic investor, stands to realize a sizable mark-up, as does Skype co-founder Jaan Tallinn and other seed-stage backers. Wired notes the offering could mint a wave of new millionaires across Anthropic's San Francisco employee base. On the customer side, enterprises betting on Claude — particularly Claude Code, widely considered best-in-class — gain a more durable counterparty. Competitively, the filing pressures OpenAI, which Wired reports is targeting a September IPO of its own.
What to watch next. The first public S-1 amendment will expose actual revenue recognition, cloud-compute cost structure, and the governance role of Anthropic's Long-Term Benefit Trust — a structure Wired flags as a possible drag on pricing. Watch also for resolution of Anthropic's lawsuit against the Pentagon over the Hegseth-era supply-chain sanctions, which executives have said threaten billions in federal sales.
Sources:
- [HN · 501↑] Anthropic confidentially submits draft S-1 to the SEC — Hacker News
- Anthropic has officially filed to go public — The Verge AI
- Anthropic, now atop the AI bubble, files for its IPO — The Register AI
- Anthropic files to go public — TechCrunch AI
- Anthropic Confidentially Files for What Could Be the Largest IPO Ever — Wired
2. Nvidia's RTX Spark puts an Arm CPU at the center of Windows — with Microsoft's Surface Laptop Ultra leading

What happened. At Computex in Taipei, Nvidia introduced RTX Spark, a 1-petaflop Arm-based PC platform pairing a 20-core Nvidia Grace CPU co-developed with MediaTek, a Blackwell GPU with up to 6,144 CUDA cores and fifth-generation Tensor Cores, and up to 128GB of unified LPDDR5x memory connected via NVLink-C2C. Shipping partners this fall include Asus, Dell, HP, Lenovo, MSI, Microsoft, with Acer and Gigabyte to follow. Microsoft's flagship implementation is the Surface Laptop Ultra: a 15-inch, sub-4.5-pound machine with a 2880×1920 mini-LED panel at 2,000 nits, a replaceable SSD, and a full I/O complement including HDMI, USB-A, USB-C, and SD. Pricing was not disclosed.
Why it matters. This is Nvidia's most credible run at the Windows client CPU market, a category Jensen Huang has now sized at $200 billion. As The Verge frames it, the launch is positioned as Windows' M1 moment — an attempt to deliver the performance-per-watt envelope Apple has held since 2020, but with native CUDA and a GPU stack that Qualcomm's Snapdragon X cannot match. Ars Technica notes that prior Nvidia forays into Windows client silicon (the Tegra-era Surface RT) ended in a $900 million Microsoft write-down; this attempt arrives with a substantially deeper software stack, including a Microsoft-Nvidia co-developed sandbox for local AI agents and an optimized Prism emulation layer with AVX/AVX2 support.
Who is affected. Intel, AMD, and Qualcomm face their first three-way contested fight for premium Windows sockets in over a decade. ISVs are being pulled along: Adobe is rearchitecting Premiere and Photoshop for the platform, Riot and KRAFTON are bringing native Valorant, League of Legends, and PUBG builds, and over 100 Windows software vendors have signed on, per TechCrunch. For buyers, the immediate question is pricing — Nvidia's developer-targeted DGX Spark sells for roughly $4,800, suggesting consumer configurations will skew premium.
What to watch next. Pricing, battery life, and Prism emulation performance on real-world x86 workloads will determine whether RTX Spark stays a workstation-class niche or pressures the mainstream. Watch Qualcomm's response, the rollout cadence at Dell's XPS 16 Creator Edition and Lenovo's Yoga Pro 9n, and whether Apple's M-series roadmap accelerates.
Sources:
- Nvidia chases $200B CPU market with AI agent PCs from Microsoft, Dell, and HP — TechCrunch AI
- [HN · 230↑] Microsoft builds MacBook Pro rival with NVIDIA-powered Surface Laptop Ultra — Hacker News
- This could be Windows' M1 moment — but expect it to cost a ton — The Verge AI
- Nvidia RTX Spark comes to Windows PCs with Arm CPU, RTX GPU, and unified memory — Ars Technica
- Microsoft's Surface Laptop Ultra looks like its first true MacBook Pro competitor — Ars Technica
3. OpenAI breaks ground on Stargate Michigan as its models go GA on AWS Bedrock

What happened. OpenAI broke ground Monday on "The Barn," a 1GW Stargate data center campus in Saline, Michigan, alongside Governor Gretchen Whitmer and partners Oracle, Related Digital, Walbridge, and Blackstone. The company committed that local ratepayers will not subsidize the project's energy infrastructure, that cooling will use a closed-loop system with water consumption comparable to an office building, and that construction will be union-built — projected at 2,500 union construction jobs, 450 permanent onsite roles, and 1,000 indirect jobs. The project is forecast to generate $1 billion in tax revenue over its lease term. OpenAI is also committing up to $45 million in Codex credits to more than 400,000 Michigan post-secondary students for the 2026–2027 academic year.
Separately and on the same day, OpenAI announced general availability of its frontier models and Codex on AWS, delivered through Amazon Bedrock in both Commercial and GovCloud regions. Codex, which OpenAI says is used by more than 5 million people weekly, is now accessible inside AWS-native security, governance, and procurement workflows.
Why it matters. The two announcements operate on different timescales but reinforce each other. Stargate Michigan extends the physical compute footprint OpenAI began in Abilene, Texas, addressing what Sam Altman told CNBC is now coding-driven demand outpacing supply. The AWS deal addresses the inverse problem — distribution. Bedrock availability removes the procurement, billing, and compliance friction that has kept many regulated enterprises from putting OpenAI models into production, particularly in government environments where GovCloud certification is a prerequisite.
Who is affected. AWS gains a marquee model partner alongside Anthropic, blunting Azure's exclusivity narrative around OpenAI. Enterprise buyers — especially in regulated sectors and federal agencies — now have a sanctioned procurement path. For Michigan, the project signals reindustrialization tied to AI infrastructure, with NABTU as the labor partner. Oracle and Related Digital deepen their roles in OpenAI's build-out.
What to watch next. The forthcoming Daybreak release — OpenAI's cyber and Codex Security suite — is slated to arrive on AWS, and its uptake among security teams will be an early indicator of whether the Bedrock channel converts to material revenue. On the infrastructure side, watch power-interconnect timelines in Saline and the next Stargate site announcements.
Sources:
- OpenAI frontier models and Codex are now available on AWS — OpenAI Blog
- Building the infrastructure for the Intelligence Age in Michigan — OpenAI Blog
- An interview with Sam Altman on OpenAI's massive Stargate data center project in Saline, Michigan, coding models being the biggest driver of AI demand, and more (CNBC) — Techmeme
- The Texas Town at the forefront of OpenAI's Stargate Project — YouTube · OpenAI
- [HN · 290↑] OpenAI frontier models and Codex are now available on AWS — Hacker News
Monday's three threads pull in the same direction: the AI industry is industrializing simultaneously across capital, silicon, and distribution. Anthropic's filing tests whether public markets will fund frontier R&D at near-trillion-dollar valuations; Nvidia's RTX Spark extends the AI compute stack from the data center to the laptop; and OpenAI's parallel Michigan and AWS moves show that physical capacity and enterprise channel access are now considered a single problem. The common assumption underneath all three is that demand for inference and agentic compute will keep compounding — a thesis Anthropic's eventual public financials will be the first to put on the record.

