Everyone hates AI now

OpenAI filed confidentially for an IPO, joining SpaceX and Anthropic in the public-market pipeline, as the AI industry faces growing public skepticism. The filing raises questions about revenue recognition, gross margins, and capital efficiency, while SpaceX is expected to price its IPO at a $1.7 trillion valuation.

Everyone hates AI now Great time to go public, OpenAI Tuesday. No, we’re not talking about AI all newsletter. Yes, we get into OpenAI’s IPO filing, but we’re also checking out Zepto’s debut, the latest Databricks round, the scale of capital flying around tech today, and we close with a few notes on why everyone hates AI now. To work — Alex 📈 Trending Up Japanese AI https://x.com/liquidai/status/2063095523318415744 … crypto unicorns https://fortune.com/2026/06/09/morpho-fundraise-a16z-crypto-paradigm-ribbit-capital-175-million/ … Apple, the model company https://machinelearning.apple.com/research/introducing-third-generation-of-apple-foundation-models … getting booed https://www.nbcnews.com/politics/donald-trump/trump-booed-madison-square-garden-game-3-nba-finals-rcna348969 … AI speed https://mimo.xiaomi.com/blog/mimo-tilert-1000tps … Finnish decacorns https://resiliencemedia.co/iceye-the-finnish-satellite-startup-nabs-e1b-at-a-e10b-valuation-amid-growing-demand-for-space-intel/ … NATO innovation https://sifted.eu/articles/alta-ares-defence-tech-50m-round/ … Amazon’s Starlink competitor https://www.pcmag.com/news/fcc-gives-amazon-leo-a-reprieve-on-satellite-launch-deadline-with-one-condition … undersea compute https://www.theguardian.com/world/2026/jun/09/worlds-first-wind-powered-underwater-datacentre-starts-operating-in-china … ever-larger AI models https://www.reddit.com/r/accelerate/comments/1u170yp/claude fable 5 claude mythos 5 and gpt56 pro will/ … the Axis of Derp https://www.bbc.com/news/live/cdepg4kw985t … OpenAI’s IPO prospects: The company filed confidentially to go public https://openai.com/index/openai-submits-confidential-s-1/ , it announced yesterday. News that the ChatGPT maker intends to list is not a surprise; that we now have SpaceX, Anthropic, and OpenAI all in the public-market hopper is beyond exciting. When Anthropic recently filed, we had five questions https://www.cautiousoptimism.news/five-questions-for-ais-biggest-ipo/ . Each query applies neatly to OpenAI with only modest editing: How does OpenAI count revenue sourced from cloud partners; what’s the gross margin profile of AI inference; how much money did OpenAI lose offering underpriced subscription plans; how capital-efficient has OpenAI been in terms of compute capex; does OpenAI look good under normal accounting methods? We have other questions, especially around the company’s revenue mix, but in terms of the information we need to answer ‘ are these AI labs good businesses ‘, our core queries are contentedly similar no matter which AI lab we’re examining. - In the military, you hurry up to wait. In IPO-land, a confidential filing is roughly as useful as no filing at all for what we can learn. When we do get the numbers, expect this blog to lose its mind respectfully for a few days. - If you add up SpaceX’s target IPO valuation and the most recent two private-market valuations of AI labs, you get $3.5 trillion. That’s more than Apple, Alphabet, Amazon, Meta, and Microsoft were worth https://techcrunch.com/2017/07/19/techs-5-biggest-players-now-worth-3-trillion/ . Then again, that was nine years ago combined in 2017 SpaceX? SpaceX is pipped to price its IPO on Thursday and begin trading on Friday. At its $135 per-share IPO target price, the space launch, connectivity, AI, neocloud, and social media company is worth $1.7 trillion on a fully diluted basis minus 1.3 billion shares held by https://www.renaissancecapital.com/IPO-Center/News/119552/SpaceX-sets-terms-for-record-breaking-$75-billion-IPO Musk that have yet to vest https://www.renaissancecapital.com/IPO-Center/News/119552/SpaceX-sets-terms-for-record-breaking-$75-billion-IPO . Not everyone agrees on the price. Recall that Morningstar is in the market, making friends and influencing people https://www.cautiousoptimism.news/microsoft-i-owe-you-an-apology-i-wasnt-familiar-with-your-game/ by arguing that SpaceX is worth around $781 billion. Other market watchers are coming to similar conclusions https://www.decodingdiscontinuity.com/p/spacex-ipo-why-enterprise-ai-800m-black-hole . Some analysts https://www.linkedin.com/pulse/spacex-spcx-going-boldly-where-one-has-gone-before-david-trainer-9hste/ are flagging concerns about SpaceX’s internal controls and its by one read declining profitability. Some well-known investors don’t like “ how capital-intensive SpaceX has become https://www.businessinsider.com/spacex-ipo-steve-eisman-big-short-elon-musk-ai-spcx-2026-6 .” I have no hard view on what SpaceX is worth, because the market doesn’t value Musk’s companies along calculable lines. That’s fine, but it makes prognostication difficult. Am I hunting for IPO allocation? Nope, I’ll hold onto my index funds and implicit stakes in Anthropic https://www.cautiousoptimism.news/good-news-you-already-own-anthropic-shares/ , OpenAI, and, via Alphabet shares, SpaceX and probably do just fine. - Naturally, the banks, with their beaks in the wetting trough, are bullish https://www.reuters.com/legal/government/wall-street-banks-ceos-promote-spacex-in-flashy-events-with-bofa-morgan-stanley-2026-06-04/ . I am curious: What do Are you going to buy shares in the IPO? Hit reply and let me know — if anyone is particularly trenchant or humorous, I may share your missive in the newsletter. you think a fair IPO valuation for SpaceX would be? The insane amount of money being thrown around in tech today: Databricks is raising private capital again, pushing off its IPO even further. Our joke that by the year 3812, the data and AI company would find itself busy raising a Series Z-4 may prove prescient. We’re making good progress towards the goal. The new Databricks round, The Information reports, will value the company “between $165 billion and $175 billion,” in what will be its Series M. I have never heard of such a thing. At some point, I think Ali Ghodsi should just admit that he never wants to go public https://www.youtube.com/shorts/NsLKQTh-Bqo , and won’t - Databricks most recently reported https://www.databricks.com/company/newsroom/press-releases/databricks-grows-65-yoy-surpasses-5-4-billion-revenue-run-rate that it closed 2025 with a run rate exceeding $5.4 billion, positive free cash flow in the preceding 12 months, a $1.4 billion run rate for its AI products, and insane net retention. All very impressive - The median market cap of a company in the S&P 500, by the by, is ~$42 billion https://www.fidelity.com/learning-center/trading-investing/markets-sectors/why-market-cap-matters . Turning the page, Cursor’s revenue has scaled to a $4 billion run rate https://www.forbes.com/sites/richardnieva/2026/06/08/cursor-4-billion-annualized-revenue/ , up from $3 billion in April and $2 billion in February; that’s a doubling in a little over a quarter. We can expect Elon Musk to buy Cursor after SpaceX lists for the previously agreed-upon price of $60 billion, though I suspect that Cursor’s investors will be bummed to see it sell for so little ; weird times. The money game is afoot everywhere. Over in China, Moonshot is raising billions more https://www.bloomberg.com/news/articles/2026-06-08/china-s-moonshot-ai-seeks-30-billion-value-in-new-funding-talks?embedded-checkout=true again , and the nation itself is turning over a $295 billion play to build more AI compute inside its borders https://www.bloomberg.com/news/articles/2026-06-09/china-prepares-295-billion-plan-to-fund-nationwide-ai-buildout . The numbers are insane; the revenue, the growth, the investment, the valuations; everyone is getting rich but you - I have lost all grounding on numbers. I recall when Instagram sold for $1 billion, an event that was treated like the Second Coming. Now that historic exit is the revenue leading startups are racking up in mere months. Wild. 📉 https://finance.yahoo.com/news/servicenow-pledges-1-5bn-investment-110000403.html Trending Down 📉 https://finance.yahoo.com/news/servicenow-pledges-1-5bn-investment-110000403.html Doing your own thinking https://techcrunch.com/2026/06/08/notebooklms-new-update-will-help-you-build-source-repository-from-chat/ … AI spending https://www.wsj.com/pro/private-equity/accel-backs-startup-pointfives-ambition-to-help-rein-in-runaway-ai-costs-77d63439?st=7K5Wgf&reflink=desktopwebshare permalink ? … non-corrupt business activity https://www.yacnews.com/albania-is-not-for-sale-kushners-4-billion-resort-triggers-flamingo-revolution-asset-freeze-and-an-eu-warning/ … staying private forever https://www.cnbc.com/2026/06/09/perplexity-ipo-2028-as-anthropic-openai-prepare-listings.html … AI costs https://9to5google.com/2026/06/08/google-ai-plus-price-drop/ ? … India-Starlink relations https://www.bloomberg.com/news/articles/2026-06-09/starlink-india-launch-hits-security-roadblock-before-spacex-ipo?accessToken=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzb3VyY2UiOiJTdWJzY3JpYmVyR2lmdGVkQXJ0aWNsZSIsImlhdCI6MTc4MTAxNDI3NiwiZXhwIjoxNzgxNjE5MDc2LCJhcnRpY2xlSWQiOiJUR0NJTTBSMjRVOEIwMCIsImJjb25uZWN0SWQiOiJCNzg3N0QzQzg1QzM0MTAwOEEwNUUwNjdBQzNFMzc5MSJ9.AXsl3T4VQC1enmWF6NcjDUNqtlwuKkk-ml6hKkiZuSY … AI costs redux https://thenewstack.io/anthropic-claude-cowork-promotion/ ? … Chinese rare earth dominance https://www.wsj.com/world/china/a-new-front-is-opening-in-the-fight-to-break-chinas-rare-earth-dominance-8b96729d ? … free speech https://www.bloomberg.com/news/articles/2026-06-08/trump-will-ask-supreme-court-to-revive-his-475-million-cnn-suit … joy at CBS News https://www.semafor.com/article/06/07/2026/cbs-news-in-the-crossfire …… Apple shares after Siri updates Zepto’s cash burn as it goes public: Don’t think that it’s only the United States enjoying the IPO game. Over in India, Zepto is going public, too. The quick delivery business updated its IPO filing here https://cdn.zeptonow.com/investors/documents/Zepto%20Limited%20-%20UDRHP-I.pdf , detailing its first-quarter performance: Converting those figures in millions of rupees into dollars, Zepto’s Q1 operating revenue soared from $534.3 million in 2024 to $1.30 billion in 2025 +194.4% to $2.42 billion in its most recent quarter +104% . Losses rose from $549.9 million to $631.7 million from Q1 2025 to Q1 2026, but cash burn declined over the same period. In percentage-of-revenue terms, Zepto is making some progress towards profitability. Last valued at $7 billion, Zepto intends to raise about $840 million in its IPO. TechCrunch notes that Zepto’s advertising business is a major growth driver https://techcrunch.com/2026/06/08/zeptos-ipo-filing-reveals-fast-growth-bigger-losses-and-a-valuation-question-nobodys-answered-yet/ for its parent company; we’ll bring you more when it begins trading. - Zepto is an Indian company, but it raised capital from General Catalyst, StepStone, Y Combinator, and other American investors. Its exit is a truly global affair. And one that could bring much-needed liquidity to private market players. As much as SpaceX? No, but isn’t it nice to talk about a growing company that isn’t building foundation models? Everyone hates AI now The vibes in AI are split. On one hand, market excitement that a version of Anthropic’s mythical Mythos model could be released as soon as today, rising token demand https://openrouter.ai/rankings , and ample vertical AI fundraising https://techcrunch.com/2026/06/09/sandstone-raises-30m-to-bring-ai-to-in-house-legal-teams/ . On the other hand, anti-AI agitation that cuts across party lines https://www.tiktok.com/@moreperfectunion/video/7646428527037910285 . Discontent is rising https://x.com/EricNewcomer/status/2064053439114469512 . Open a poll this year, and what do you find? Anti-AI sentiment. Gallup found that seven in ten Americans https://news.gallup.com/poll/709772/americans-oppose-data-centers-area.aspx are opposed to local data center construction; Pew uncovered a spike https://www.pewresearch.org/short-reads/2026/03/12/key-findings-about-how-americans-view-artificial-intelligence/ in AI concern amongst the public starting in 2023; the same poll found that just 10% of American adults were more ‘excited than concerned’ about AI last year. YouGov found https://yougov.com/en-us/articles/54888-more-expect-artificial-intelligence-ai-affect-economy-negatively-than-positively-americans-split-how-ai-will-impact-their-lives-may-29-june-1-2026-economist-yougov-poll that just under two-thirds of Americans think AI is advancing too quickly ‘too slowly’ earned 2% of the vote . Why? Worries are pandemic that companies want to replace human workers with AI while concentrating power for themselves CBS polling https://www.cbsnews.com/news/opinion-poll-government-ai-jobs/ . And the trends are bad . A recent poll of New Hampshire residents hardly the epicenter of AI progress, I know found that between October 2025 and May 2026, consumer sentiment declined when adults were asked to estimate the impact AI will have on the economy and their lives. Observe the shifts: At the same time, more New Hampshire residents reported using AI in May of this year than last October. What’s going on? A lot, but I think we can outline a few major drivers of negative AI sentiment: The companies all-in on AI are cutting staff: While it appears that we’re not heading for a near-term white-collar jobs apocalypse, early-career workers are shouting that the job market is a mess. Arguments that AI will drive job growth are being undercut by the endless headline reel of major tech companies trimming or slashing staffing levels. Indeed, I think there’s a stronger positive correlation between rising AI usage and reduced human headcount than the opposite, on a total-market net basis. Voter concer n: AI will not make it easier for you to support your family and may make it harder for you to hold onto your job. Tech companies are operating on a different timetable than the public: Why are data centers so unpopular? Tech folks like to argue that data centers support the construction industry and create some long-term jobs in the communities where they are located. Sure. But from a local perspective, data center companies browbeating planning commissions into approving their installations is more than just eating up all the goodwill that a few new jobs create. Communities feeling utterly crushed, misled, or simply left in the dark by outside economic interests are not rare at all https://www.foxcarolina.com/2026/05/01/spartanburg-co-residents-say-they-were-misled-about-data-center-project/?outputType=amp . Here’s more https://theindiopost.com/data-center-town-hall-turns-contentious-as-residents-from-across-the-valley-demand-answers/ , and more https://www.cleveland19.com/2026/04/14/perry-village-residents-demand-answers-proposed-data-center/ , and more https://www.stlpr.org/health-science-environment/2025-08-15/angry-st-charles-residents-question-secretive-planned-data-center . And more https://www.wcmu.org/local-regional-news/2025-12-02/saline-residents-protest-data-center-project-chant-no-secret-deals . Voter concern: C ompanies are invading local areas with little regard for the people who already live there, often following technology-friendly timelines for data center development that don’t mesh with local culture. The gains from AI are concentrated, while dislocation is distributed: Once you get that shiny new data center in your community, what do you get for it? The median voter would probably say higher energy prices and less confidence in long-term water supplies, regardless of what you or I might think of such concerns. The gains? Concentrated inside already-wealthy companies, AI labs that raised money from the already-rich, and business leaders who can use AI to replace some human staff. So a small town may watch itself get transformed from a rural community into a quasi-industrial zone, with little of the economic upside generated staying local. Voter concern: The American right doesn’t like tech companies after their leaders shouted for years that they were anti-conservative; the American left doesn’t like tech companies after watching tech leaders cozy up to Trump in his second term. Why would voters want to help those companies by letting them build massive data centers down the road from cows? Workplace AI usage is a mess: We’ve covered how the corporate view of AI has rapidly evolved. From ‘use AI or get fired’ to ‘use too much AI and we may have to fire you’ is one hell of a shift in mere quarters. Workers were annoyed at being told to use AI implicitly stating that their work was mere proto-automation, that they were the biomechanical robots of the business world , and are now annoyed at being told to use less of it to prevent rising costs; we told you , the refrain goes. Voter concern: If I use it, will it fire me? Or, if I use it, will I get fired? Either way, AI is not engendering confidence in the labor market. And most people need their job to pay their rent or mortgage. Uncertainty about your ability to stay housed is an electoral loser. Anti-AI boomer slop: 404Media reports https://www.404media.co/ai-grifters-are-making-anti-data-center-slop-with-ai/ that over on the original Meta social network a land that you and I rarely, if ever, visit , “hundreds of anti-data center Facebook pages churning out AI-generated slopaganda” against data centers. Yes, grifters are using AI to generate anti-AI slop, firing up the same GPUs that they are protesting against. Ironic, but market incentives appear to reward the activity. Voter concern: Older people love Facebook; they are eating from a trough of posts arguing that farms matter more than data centers. Result? The most reliable voters, and often some of the most conservative, are being anti-AI-pilled by AI. Not good for AI polling. Tech companies appear to understand that something needs to change. OpenAI is making big noise about empowering everyone https://openai.com/index/built-to-benefit-everyone-our-plan/ , Meta wants to invest in training more data center workers https://www.wsj.com/tech/ai/meta-launches-workforce-academy-to-train-workers-to-build-data-centers-35470a80 , the list goes on. But the promises and actions are spit on hot metal; all sizzle, no real temperature change. The most recent idea is that the major AI labs should become partially nationalized. This is a rare Bernie Sanders-President Trump point of agreement. If the United States government buys, or merely takes a stake in the Anthropics of the world, then all is well, right? The public gets some upside, and thus the equation equals out? Not even close. Let’s say the government got a cut of all major AI companies, worth $1 trillion all in. What would the impact be on the voting public? Very little, while all the above concerns would still hold true. No, the only real way to reduce AI angst is to tie data center construction and operation to local coffers, perhaps by paying a percentage of net revenue to local government arts, transit, healthcare, policing . But that would cost real money, undermining the economics of already-intensely expensive GPU buildouts. I doubt there’s real appetite for such an act. So where are we left? In the middle of a buildout gold rush that is throwing dirt all over communities that, until recently, had little or nothing to do with technology. I remain an AI bull, but one sympathetic to the concerns of others this precludes me from the current technology in-crowd; I’ll survive . This means I think local communities should have a say in what happens in their neighborhoods, even if giving them that obvious right may slow overall data center construction; there are enough incentives pushing it, and we can afford to apply the brakes a little today to allow for an even more computationally heavy future. Why did we frame our bullets in terms of voter concerns? Because we’re heading into a period with both national elections and major AI IPOs. What could possibly go wrong?