# 本月生成式AI失去了魔力 - 来源：Gary Marcus：The Road to AI We Can Trust（RSS） - 作者：Gary Marcus - 发布时间：2026-06-27 06:43 - AIHOT 分数：48 - AIHOT 链接：https://aihot.virxact.com/items/cmqwtsmeo002tslp0957l26f1 - 原文链接：https://garymarcus.substack.com/p/the-month-generative-ai-lost-its ## AI 摘要 OpenAI倾向于推迟IPO至明年，因估值不及预期且零售投资者兴趣不足。SpaceX一周跌11.74%，Nvidia近一月跌超8%，Oracle跌22%，CoreWeave跌4%，Microsoft跌10%，SoftBank跌12%，Cerebras跌32%。美国AI政策混乱遭广泛批评，中国模型快速崛起。作者认为大语言模型正商品化。 ## 正文 The month Generative AI lost its mojo June is not over, and anything could still happen, but a lot already has. Gary Marcus Jun 26, 2026 “WeWork’s IPO fell apart. I’ve been calling OpenAI the (possible) WeWork of AI for a long time. I will not be surprised if their IPO plans fall apart, too” – your author, June 10 News, mostly just from the last few days: OpenAI is leaning towards delaying its IPO until next year: The New York Times@nytimes OpenAI is leaning toward holding off its IPO until next year, people involved in the company’s deliberations said, a delay that highlights the unclear paths of A.I. titans. nyti.ms OpenAI Leans Toward Holding Up I.P.O. Until Next Year 9:11 PM · Jun 25, 2026 · 199K Views 42 Replies · 88 Reposts · 314 Likes One rumor is that Altman was told he can’t get the valuation he wants: NIK@ns123abc 🚨BREAKING: OPENAI IPO DELAYED Altman told everyone OpenAI is worth a TRILLION dollars His own advisors warned him that retail investors aren’t buying it… gave him 2 options: >wait until 2027 for $1T valuation >or lower valuation for 2026 IPO Altman: any valuation under $1 10:37 PM · Jun 25, 2026 · 187K Views 88 Replies · 102 Reposts · 1.06K Likes Another rumor is that Altman is worried that the SpaceX IPO isn’t taking off. Also, per the Times, “….OpenAI’s advisers, in conversations with the company over the past week, …caution that it may not find much enthusiasm from retail investors for its own shares, two of the people involved said”. Still others have wondered whether OpenAI’s financials are solid enough yet. However you slice it, the delay reflects a lack of confidence. As I long predicted, troubles with OpenAI may hurt other companies that they do business with. We are perhaps already seeing evidence for that SpaceX, increasingly pitched as an AI play, has lost steam (up slightly for the day, but down 11.74% for the week), and its bonds are doing badly, too: *Walter Bloomberg@DeItaone $SPCX - SPACEX BOND SELLOFF DEEPENS SpaceX's $25 billion bond sale is suffering unusually steep losses, with paper losses exceeding $300 million. Traders say fast-money investors may be exiting, while concerns over the company's valuation, long-term cash flow and heavy 12:25 PM · Jun 26, 2026 · 219K Views 102 Replies · 138 Reposts · 858 Likes Nvidia is down over 8% over the last month; Oracle is down about 22% over the last month, CoreWeave is down a bit over 4% for the month, Microsoft 10% over the last month. SoftBank down about 12%. Chip company Cerebras is down 32% over the last month. See yesterday’s essay on the Fizzle US AI policy is in shambles, drawing widespread scorn: Zvi Mowshowitz@TheZvi Our new AI policy is that the White House decides ad hoc, for whatever reasons it likes, who does and does not get access to frontier intelligence. This seems rather maximally terrible. Andrew Curran @AndrewCurran_ The US Government has requested a slow staggered rollout of GPT-5.6, and OpenAI has agreed. During this phase the government will approve each user individually. This will probably be the norm for all frontier models from all labs from now on. 10:19 PM · Jun 25, 2026 · 289K Views 73 Replies · 141 Reposts · 1.39K Likes Deirdre Bosa@dee_bosa Washington is doing more for Chinese AI than Beijing ever could First Fable now GPT 5.6 Stephanie Palazzolo @steph_palazzolo New w/ @leomschwartz @amir: The Trump admin has asked OpenAI to stagger the release of GPT-5.6 over security concerns. On Thursday, CEO Sam Altman told staff that the government will be approving access to GPT-5.6 customer by customer, a highly unusual approach. 12:35 PM · Jun 26, 2026 · 15.1K Views 23 Replies · 15 Reposts · 225 Likes Dean W. Ball@deanwball In a matter of weeks, U.S. federal AI policy has gone from implausibly libertarian to increasingly draconian and opaque. Today, over 35 distinct observations, I analyze how we got here and offer the most succinct statement I can about what exactly I propose we should do next. 11:51 AM · Jun 26, 2026 · 46.4K Views 19 Replies · 74 Reposts · 372 Likes • Medical AI bull Eric Topol just posted this: We stress tested many frontier AI models for multimodal medical reasoning (including GPT-5, Claude 3.5, Gemini 2.5 Pro). They’re not ready. Faulty reasoning, use of inappropriate shortcuts, hallucinations. Published today @NatureMedicine nature.com/articles/s4159… 9:19 AM · Jun 26, 2026 · 69.4K Views 82 Replies · 190 Reposts · 662 Likes and there is this from Nobel Laureate Jennifer Doudna jeffrey lee funk@jeffreyleefunk Although the tech bros and some pharma think AI will cure cancer, Nobel Laureate Jennifer Doudna, inventor of the Crispr gene-editing technology says that innovation is still really in the domain of human beings right now. I’m not seeing chatbots coming up with a brand-new idea” 10:12 AM · Jun 26, 2026 · 2.45K Views 1 Reply · 8 Reposts · 23 Likes • Backlash against AI is getting really intense for many reasons, as discussed for example yesterday by Paul Krugman: Paul Krugman Why Does Everyone Hate AI? Read more 2 days ago · 2328 likes · 613 comments · Paul Krugman • Chinese models are gaining traction, fast. zerohedge@zerohedge "the share of tokens used for US models on OpenRouter has collapsed": Bloomberg 11:38 PM · Jun 25, 2026 · 385K Views 131 Replies · 373 Reposts · 2.67K Likes As I have said before, over and over, LLMs are becoming a commodity. None of this bodes well for the US AI industry. § AI bulls of course will disagree. As Noah Smith pointed out, drawing on an excellent new report from Azeem Azhar et al, revenue is clearly up But Smith’s conclusion that there is no bubble doesn’t follow. The chart he shared shows that revenue is up, not that profitability is , nor that it can be sustained, as I noted on X: Gary Marcus@GaryMarcus Chart of increasing revenue, pre the decline of token-maxxing, not looking at the immense costs/lack of profitability, not considering threats from open-source models, price-war dynamics etc. @noahopinion, I think you are hasty in your conclusions here. Noah Smith 🐇🇺🇸🇺🇦🇹🇼 @Noahpinion "AI is a bubble" is looking less and less likely by the day 8:09 PM · Jun 25, 2026 · 45.7K Views 11 Replies · 6 Reposts · 112 Likes Smith fired back, pointing to what might be Anthropic’s first profitable quarter. Noah Smith 🐇🇺🇸🇺🇦🇹🇼@Noahpinion "Lack of profitability" Gary Marcus @GaryMarcus Chart of increasing revenue, pre the decline of token-maxxing, not looking at the immense costs/lack of profitability, not considering threats from open-source models, price-war dynamics etc. @noahopinion, I think you are hasty in your conclusions here. 10:23 PM · Jun 25, 2026 · 53.4K Views 22 Replies · 6 Reposts · 199 Likes But he left a lot out. In focusing on Anthropic’s not yet confirmed spring quarter Smith ignored the fact that the spring quarter — which he selected post hoc and which looks different from all others — was (a) during peak tokenmaxxing, (b) before recent Chinese advances, and (c) in a quarter in which SpaceX gave Anthropic a big one-time subsidy, quite possibly larger than the profit. And, again, the WSJ prediction of a positive quarter that Smith pointed to was speculation; it’s not at all clear that it will actually come to pass, given all that happened since then. With the decline of tokenmaxxing, the rise of Chinese models, and the U.S. government pause on Fable, June was not as kind to Anthropic as April and May were. So we’ll see. Ed Elson’s overall take on Anthropic’s profitability, just posted, was this: Ed Elson@edels0n Anthropic is the closest thing AI has to a profitable lab. But you still get an operating loss of $11 billion. Can AI actually make money? (Spoiler: so far, no.) 👇 4:15 PM · Jun 26, 2026 · 5.93K Views 8 Replies · 7 Reposts · 57 Likes Even Polymarket sees it now; the game has changed: Polymarket@Polymarket JUST IN: Businesses are shifting from AI “tokenmaxxing” to efficiency, threatening the explosive growth of OpenAI & Anthropic. 2:26 PM · Jun 26, 2026 · 114K Views 101 Replies · 69 Reposts · 1.01K Likes § I tried to distill my basic argument yesterday in the FT. Here’s one part of it: If the argument I laid out there is remotely correct, hyperscaling could prove to be among the biggest financial blunders in history. P.S. An analogy worth considering: P.P.S. Fun new interview: Prof G Markets@ProfGMarkets 'AI Is Dumber Than Investors Think — ft. Gary Marcus' Ed Elson is joined by Gary Marcus to discuss why he’s concerned about the fact that we’re all-in on AI. They explore why he argues generative AI is inherently unreliable, whether the concerns surrounding Anthropic's Mythos 11:03 AM · Jun 26, 2026 · 929 Views 1 Reply · 3 Reposts · 11 Likes Discussion about this post No posts Ready for more?