Anthropic Tortoise laps OpenAI Hare, for now. ARD #75
...Anthropic revenues $50B+ rate & pre-IPO raise at ~ trillion $ valuation
The frame running through today’s episode: Anthropic’s product and revenue momentum has been building for a while. For most of the last three years, OpenAI was the consensus AI Tech Wave front-runner by every public metric — ChatGPT user counts, brand mindshare, Microsoft partnership, fundraising scale. Anthropic was the deliberate second mover. That’s changing.
Today’s WSJ report — Anthropic revenues will reach $50 billion by June — moves it ahead of OpenAI on the metric Wall Street watches into both companies’ trillion-dollar IPO window by year-end. The Pepsi is turning into Coke, as MP put it on-air. The race is now tight and Anthropic has been the proverbial hare to OpenAI’s turtle.
Here’s the broader frame: when the media and investors look at this, they tend to look at it as a zero-sum game. The AI Tech Wave is not a zero-sum game. We’re four years into the post-ChatGPT moment for chatbots, now onto AI reasoning and AI agents along the long road on AI functionality — all the way to that mythical AGI in the distance that AI gurus debate about. This race continues even way past the IPOs. The IPOs are just a beginning point, not an end point.
Three Takes today on three concurrent moves driving Anthropic’s lead — revenue crossover, near-trillion-dollar private valuation, and tactical Stainless acquisition. Plus a Gadget AI on Meta’s new Incognito Meta AI Chat on WhatsApp — and two reader questions on what gets mainstream traction next:
(1) Anthropic Crosses OpenAI in Revenues. WSJ reports that Anthropic revenues will reach $50 billion by June — moving it ahead of OpenAI on the metric Wall Street watches ahead of both companies’ trillion-dollar IPOs. Last week Dario Amodei said it was a 10x pace every few months, and with more data center capacity — more compute — he said the run rate could grow 50x, 70x, 80x. (Not sustainable at those numbers, but the direction is clear.) June is also when SpaceX/xAI is planning to get its public-market listing if Elon can launch it. Anthropic’s path to the lead has been its single-minded focus on the Coding market anchored in Enterprise distribution — Claude Code, the Claude Cowork product layer, and a developer-tools API ecosystem that became the default for a lot of the high-end coding workload. OpenAI is now visibly trimming its AI applications portfolio to focus back on Anthropic in Enterprise and Coding — and OpenAI’s Codex is not too far behind, aggressively ramping revenue as well. A reminder that OpenAI has just finished renegotiating its agreement with Microsoft — the $13 billion of initial capital that gave them the early oomph and headroom to race ahead of Anthropic. That’s now behind them, and they’re running hard. The reporting: WSJ — Anthropic Was Behind, Now It’s the AI Boom’s Front-Runner. Standing backcat — Coke-vs-Pepsi framing: AI-RTZ #887 — OpenAI vs Anthropic, ‘Coke vs Pepsi’. And: AI-RTZ #1037 — OpenAI Trims AI Applications, Focuses on Anthropic in Enterprise and Coding.
MP Take: This lead is driven by Anthropic’s single-minded focus on the Coding market around the Enterprise with terrific Product Market Fit (PMF) at global scale. These new resources mean Anthropic is now neck and neck with OpenAI through the IPO. The market for now is big enough for both. Still early days in this AI Tech Wave.
(2) Anthropic Raises More Private Money — Close to a Trillion. The NYTimes reports Anthropic raising additional private funding at a valuation approaching $950 billion — a near-trillion-dollar private mark that puts it in the same orbit as OpenAI and xAI ahead of the IPO window. Both companies now have enough momentum and investor enthusiasm around the world to see valuations of a trillion dollars. But these are nominal numbers — on traditional valuation metrics they’re going to be at multiple-multiples of revenue. These companies are not going to show profits anytime soon, given they’re continuing to aggressively build multi-gigawatts of data centers. Every gigawatt of data center build is a $50 billion number. That’s the con to the pro. The reporting: NYTimes — Anthropic Funding at $950 Billion Valuation. Standing backcat — Siblings framing: AI-RTZ #928 — OpenAI and Anthropic, Different Siblings. And: AI-RTZ #1048 — It’s Business and Personal in OpenAI/Anthropic.
MP Take: Both companies are going to be able to raise the sums they need before and through the mega AI IPOs. Unless the macro market environment changes on a broader basis. Investors are now very much in the leaning-in phase for both AI Data Center buildouts and LLM AI scaling. And the market has accepted both companies as Coke and Pepsi. Even though the roles may switch back and forth in the near-term. This ‘race’ is anyone to take. There’s a lot of headroom, lots of innovation. Both companies have new versions of their models — Anthropic with “Mythos”, OpenAI with their 5.5 Jagged PD model and various incarnations + iterations of Codex. The race continues even way past the IPOs. The IPOs are just a beginning point, not an end point.
(3) Anthropic Buys Software Tool Company Stainless for $300M — AI API/SDK Provider Used by Google and OpenAI Too. The Information reports Anthropic in talks to buy Stainless — an AI developer-tools startup that provides API and SDK generation tooling — for approximately $300 million. Stainless is notable because it’s also used by Google and OpenAI as part of their developer-facing API stack. These tools basically allow millions of customers to access the models at Anthropic, Google, and OpenAI — software glue that holds the developer experience together across the AI Tech Wave. An acqui-hire-plus-tooling move that takes a piece of cross-frontier-lab infrastructure off the table and brings it in-house at Anthropic. Months before an IPO is one of the most opportune moments for these companies to buy smaller companies — the acquired companies get stock that will be public and worth something more than if they were to remain private. The reporting: The Information — Anthropic Talks to Buy Developer-Tools Startup Used by OpenAI and Google. Standing backcat — Building AI Tools: AI-RTZ #518 — OpenAI and Anthropic Build AI Tools. And the glue thesis: AI-RTZ #693 — Importance of AI Software Glue.
MP Take: These tactical battles will continue, and point to a range of acquisitioms and acquihires for both companies in the software data and tools arenas. These are utilitarian transactions meant to expand near-term product refinement efforts. See this in every major tech wave. This one is no exception. Expect more of these from both Anthropic and OpenAI as they race through the IPO window.
Plus: Gadget AI — Meta Tries Incognito Meta AI Chat on WhatsApp. Meta is introducing “Incognito Chat” for Meta AI inside WhatsApp — a fully end-to-end encrypted, no-history, no-training conversation channel between users and Meta AI, layered onto WhatsApp’s existing global encrypted-messaging infrastructure. The framing: a completely private way to chat with AI. The context: Zuckerberg has built up huge data center capacity and the Meta Superintelligence Lab (MSL) under Alexander Wang (acqui-hired from Scale AI). He spent billions hiring dozens of people at literally tens of millions of dollars each. Now Meta needs to ramp up usage of Meta AI by the 3.5 billion+ users across WhatsApp, Facebook, and Instagram. These are the companies with billions of consumers who are still scratching their heads about this AI thing. Meta’s pitch: We know you’re scared of AI. We know you think the data is being used to train our models — pro tip, it is. But we’ll let you use this on an incognito basis. Your data will be totally end-to-end encrypted. We won’t see it. Give it a shot. A tried-and-true trick. WhatsApp is the right global surface for the play. The reporting: WhatsApp Blog — Introducing Incognito Chat with Meta AI. And: Wired — WhatsApp Incognito Chat with Meta AI. And: The Verge — Meta AI Incognito Chats. Standing backcat — Meta’s Muse Spark powers Meta AI: AI-RTZ #1052 — Meta’s Closed Muse Spark Enters.
MP Take: This is Meta trying to get ahead vs OpenAI, Google Gemini et al on mainstream usage of its AI chatbot, as these companies step on the gas for mainstream distribution. Especially with Google Gemii’s partnership with Apple and Siri for Ai search and chatbot features. WhatsApp is a global app, and could see traction on truly incognito, fully encrypted AI chats in global markets. Good tactical move by Meta. And it’ll have to be matched if successful — by Google and Apple in their own ways. Google’s I/O developer conference is next week May 19-20. Apple’s WWDC is June 8th. Lots more to come.
Today’s AI-RTZ companion #1084 covers Nvidia’s Jensen Huang getting invited at the last minute to go with President Trump on Air Force One to China for the big trade meeting that’s going on right now. Over a dozen American CEOs have accompanied the president on this trip — Tim Cook from Apple, Meta’s contingent, a full roster of tech. Now Nvidia, because chips is a very key piece of the puzzle. See what comes out in the next couple of days.
Closing Questions —
What AI application does MP see getting near-term mainstream traction? AI as companions continue to see adoption across demographic groups. The companion category — relationship-style AI, daily-routine AI, light-context personal assistants — is the one consumer-AI use case that’s showing durable engagement curves across age cohorts and geographies. Everything from the younger folks to older folks. A complex area with a lot of pros and cons, lots of debate, lots of concerns about use by really young kids, and a lot of regulatory examination here. But this is an area where consumers have figured out this is something interesting. MP will watch this keenly. It’s still early days, and there’s a lot to resolve on the stickiness of the application.
MP Take: It’s neither the most hyped category by the industry nor the most discussed by analysts, but the usage data and the recurring patterns of mainstream behavior point to it being the first surface where AI feels indispensable rather than experimental.
What role can AI Agents play for mainstream users? Unclear for now. AI companies are trying agents on AI planning and shopping activities — calendar-style coordination, recipes, party prep, invitation lists, event organization, comparison shopping. Places where most of the AI companies are hoping they’ll have revenue streams because they’ll attach payment mechanisms. ChatGPT just announced initiatives with Stripe and other partners to try that particularly aggressively. AI Agents have gotten a lot of traction, product-market fit on the Enterprise side — especially with Cloud Code and Cloud Cowork and OpenAI Codex. But on the consumer side, it’s still a wait-and-see game.
MP Take: The AI Agent technology for consumers looks great in demosdemos; the day-in-the-life mainstream adoption signal hasn’t shown up yet. We’ll see more on this from Google at their developer I/O confrence 5/19 and Apple at their WWDC conference on 6/8.
Still a work in progress for all. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)
Shorts Clips from today’s episode
Short — Anthropic Crosses OpenAI — $50B by June WSJ reports Anthropic revenues will hit $50 billion by June 2026 — moving Anthropic ahead of OpenAI on the metric Wall Street watches into both companies’ trillion-dollar IPO window by year-end. Dario Amodei last week framed it as a 10x pace every few months — and said with more data center capacity the run-rate could grow 50x, 70x, 80x. OpenAI is visibly trimming its apps portfolio to refocus on Enterprise + Coding — recognition that the competitive center has shifted.
MP Take: This lead is driven by Anthropic’s single-minded focus on the Coding market around the Enterprise. These new resources mean Anthropic is now neck and neck with OpenAI through the IPO. The market for now is big enough for both. Still early days in this AI Tech Wave.
Short — Anthropic Buys Stainless — Acquihires Continue Anthropic in talks to buy Stainless for ~$300 million — an AI developer-tools startup that provides API + SDK generation tooling. Notable because Stainless is also used by Google and OpenAI. An acqui-hire-plus-tooling move that pulls cross-frontier-lab infrastructure off the table and brings it in-house at Anthropic. Months before an IPO is the most opportune moment for these acquisitions — acquired companies get stock that becomes public and worth more than staying private.
MP Take: These tactical battles will continue, and point to a range of acquisitioms and acquihires for both companies in the software data and tools arenas. These are utilitarian transactions meant to expand near-term product refinement efforts. See this in every major tech wave. This one is no exception.
Short — Meta’s WhatsApp Incognito AI — Privacy Play Meta launches Incognito Chat for Meta AI inside WhatsApp — fully end-to-end encrypted, no-history, no-training. A tactical move to get mainstream WhatsApp users to try Meta AI without the “all my data feeds the model” concern that’s been blocking adoption. Zuckerberg has spent billions building out Meta Superintelligence Lab under Alexander Wang (acqui-hired from Scale AI). The next play: get the 3.5 billion users across WhatsApp, Facebook, and Instagram actually using Meta AI.
MP Take: This is Meta trying to get ahead vs OpenAI, Google Gemini et al on mainstream usage of its AI chatbot, as these companies step on the gas for mainstream distribution. Especially with Google Gemii’s partnership with Apple and Siri for Ai search and chatbot features. WhatsApp is a global app, and could see traction on truly incognito, fully encrypted AI chats in global markets. Good tactical move by Meta.
Short — AI Agents — Still No Mainstream PMF Yet AI Agents for mainstream consumer users: still searching for product-market fit. Big AI companies are trying agents on planning and shopping — recipes, party prep, event coordination, comparison shopping. The technology demos well; the day-in-the-life mainstream adoption signal hasn’t shown up yet. ChatGPT just announced initiatives with Stripe and other payment partners to push the agent-with-checkout idea harder.
MP Take: AI Agents have gotten real traction on the Enterprise side — Cloud Code, Cloud Cowork at Anthropic and Codex at OpenAI. But on the consumer side it is still a wait-and-see game. AI companies are hoping payment mechanisms attached to planning + shopping agents will unlock revenue. Still early days. Worth watching, not yet worth betting on.
About AI Ramblings Daily (ARD), and AI-RTZ
Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.
AI-RTZ is the morning text — a deeper written take on one idea, published by at least 5 AM EST. Today: post #1084 — Nvidia Jensen Invited Last-Minute to US/China Trade Summit.
AI Ramblings Daily is the afternoon video + podcast — my ad hoc takes and perspective on the day’s AI issues & news flow, around 14 minutes today, with short 1-minute clips for quick topic views. Today: episode #75.
Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.
Links used in today’s show (already embedded inline above; listed here for reference)
Take 1 — Anthropic Crosses OpenAI in Revenues:
WSJ — Anthropic Was Behind, Now It’s the AI Boom’s Front-Runner
AI-RTZ #1037 — OpenAI Trims AI Applications, Focuses on Anthropic in Enterprise and Coding
Take 2 — Anthropic Raises Near-Trillion-Dollar Private Round:
Take 3 — Anthropic Buys Stainless for $300M:
Gadget AI — Meta’s WhatsApp Incognito Meta AI Chat:
Q1 + Q2 — AI Companions / AI Agents Mainstream Traction:
(No external URLs — MP’s standing thesis. Q1 anchors AI Companions as the near-term mainstream-traction category; Q2 keeps AI Agents in the “still finding PMF” bucket for consumer use.)
Companion text:
The Stainless acqui-hire is the more interesting line item here than the $50 billion run rate. The reporting says the same API and SDK tooling is in production at Anthropic, Google, and OpenAI, so pulling it in-house at roughly $300 million months before an Anthropic IPO buys two things at once: developer-experience parity inside the house, and a small piece of cross-frontier-lab infrastructure that the other two now have to replace or refactor. Acqui-hires at this size used to be a talent play. In an IPO window where every lab is also a customer of every other lab's tooling, they read more like surgical strikes on the shared glue.