Fed Hesitates on Tariffs, The New Mag 7, Death of VC, Google's Value in a Post-Search World
Speakers
Philippe Laffont
— Founder and Portfolio Manager, Coatue Management
Chamath Palihapitiya
— CEO, Social Capital
David Friedberg
— CEO, The Production Board
Jason Calacanis
— Angel Investor / Founder, LAUNCH
Summary
All-In hosts Jason Calacanis, Chamath Palihapitiya and David Friedberg are joined by Coatue founder Philippe Laffont after the May 2025 Fed meeting. They debate why the Fed held rates amid tariff uncertainty, with Laffont arguing hard data is far better than sentiment and that AI token demand matters more than tariffs, while Chamath flags subprime lender valuations and liquidity as warning signs. A long segment covers Google's roughly 8% drop after Apple's Eddy Cue said Safari search volume fell for the first time, and whether Google can manage the shift from links to AI answers. The episode closes on the end of the Magnificent 7 framing, the collapse in IPOs and M&A, and Laffont's new hybrid public-private interval fund.
- The Fed held rates at 4.25 to 4.5%; Laffont argues not cutting signals a strong economy rather than a problem.
- Laffont: consumer sentiment is very weak but spending is resilient, and April was a tariff tantrum, not a tariff crisis.
- Chamath reads subprime lender price-to-book spreads and liquidity as blinking yellow leading indicators, and calls the Fed partly political.
- Friedberg notes the UK deal leaves a 10% baseline tariff, creating a new federal revenue stream that complicates rate cuts.
- Laffont's tokens over tariffs framing: Microsoft processed 100 trillion tokens in a quarter and chips and compute remain in shortage.
- Google fell about 8% after Apple's Eddy Cue said Safari search volume declined for the first time in 20 years.
- Friedberg and Calacanis think Google can manage the transition; Chamath assumes search share falls from 99% toward 75% and says the market will price that decay; Laffont finds the stock too complicated.
- Laffont calls the Magnificent 7 framing over, favors a roughly 25-name basket including private companies, and explains his new interval fund seeded with $1 billion from the Bezos and Dell family offices.
Ideas
Strong hard data supports US equities.
The Fed holding rates is not a bearish signal: if the Fed were cutting it would be because things are bad, so no cut is evidence the economy is strong. Coatue tracks a ratio of hard news over sentiment and this is the first time the hard data is so good while sentiment is so bad. Consumer spending is remarkably resilient even though consumer sentiment is terrible, visible in Visa and Mastercard earnings and in company earnings-call commentary that April spending stayed strong even after stripping out pre-buying ahead of tariffs. Sentiment is a lagging indicator that simply tracks the market. The drawdown was a tariff correction or tariff tantrum, not a tariff crisis, because two things put a floor under it: the administration budged when markets fell, and the Fed said it would not cut to bail out equities but would step in if market liquidity stopped functioning.
Subprime lender valuations signal credit rollover.
He shows a chart of the price-to-book spread between subprime lenders, Credit Acceptance versus Capital One. Historically, when these subprime lenders' price-to-book multiples escalate and reach highs, it has portended a liquidity crisis and signalled that things are about to roll over. He reads liquidity and the credit health of the American consumer as blinking yellow lights that the Fed is choosing to ignore, and he explicitly does not believe in a Fed put, so the warning is not something a policy backstop removes.
Token demand beats tariffs; chips short.
His framing is tokens greater than tariffs. Tariffs are an emotional, temporary drag that he expects to fade over the next year or two as deals, deregulation and tax breaks offset them, while AI token consumption is going vertical: Microsoft processed 100 trillion tokens in the quarter, 50 trillion in March alone, because reasoning models require far more compute. Tokens are to an AI model what fuel is to a car. Part of the drawdown was not tariffs at all but the fear that AI was not working and had no ROI, and Microsoft's capex and token data refuted that. From every private and public company he sees there is a gigantic shortage of chips and of compute power, which is why the market is recovering. After 35 years he calls this the most exciting trend he has seen and argues it marks the beginning, not the end, of American exceptionalism.
Google can transition search gradually.
The search-click-repeat paradigm is over and human-computer interaction is shifting to chat, voice and other surfaces, but Google is well placed to manage that transition rather than be destroyed by it. Its underlying models are competitive if not the best, it already has the users, a standalone Gemini app and the product surface. Flipping search to a chat interface overnight would be wrong anyway: search advertising is a $200 billion revenue stream and an AI query costs an order of magnitude more to serve, so the rational path is a gradual migration through the AI one-box and the app. Google is an organisation built on testing and incremental change, and he is positive on its ability to respond to the shift.
Diversified Google cuts costs, earnings rise.
Search is only about 56% of Google's revenue and Cloud plus YouTube, Docs, Android and Workspace mean it is no longer a one-revenue-stream company, so falling clicks below the AI box, down 15 to 35% in various studies, is significant but manageable. Google has four or five products with one to two billion monthly users, a huge data advantage and integration so deep that Gemini already pulls his calendar and works inside Docs, and YouTube search with its transcript corpus is a secret weapon it should go all in on. If management cuts a large number of employees, brings people back to the office and takes this seriously the way Sergey Brin is, earnings go massively higher even while it spends $75 billion on infrastructure.
Market starts pricing Google's search-share decay.
Google has the best models in many domains, but the problem is that it was effectively at 99% search share and the shoe has now dropped: anybody can build a model that prices the economic value of every basis point of share shift, and that repricing was initiated this week by the Apple testimony. Management should assume share goes from 99% to roughly 75% within two years, red-team what goes wrong, and then very aggressively make Gemini the front-facing window to Google Inc, accepting the cannibalisation that requires taste and courage. The bigger risk is that the total query volume is expanding, so owning 80% of a market three times larger could be fine, but Google is nowhere in that new bucket and cannot show up in 18 months asking to be picked. Waiting for data means reacting to Apple, OpenAI and Facebook announcements, which destroys the morale of its product managers and engineers. He says the market will now start to price this decay in, while stating that he is long.
Google too complicated; possible next IBM.
Google is worth about $1.8 trillion versus roughly $300 billion for ChatGPT, six ChatGPTs, and he questions whether that is the right ratio going forward. He lived through the yellow pages being replaced by blue links, but those companies died partly because they were levered, whereas Google has no leverage, a lot of cash and three good businesses in Waymo, YouTube and Cloud. The problem is that search is roughly 60% of revenue and effectively all of the profits, a textbook innovator's dilemma in which the head of the Gemini app and the head of the search box are fighting over the same business. He has not formed an opinion, says these stocks are a little too complicated for him and that sometimes you have to admit something is just tricky, and his open question is whether Google becomes the next IBM that survives a long time with little growth or genuinely re-engineers itself. He does add that Google is right to keep investing rather than harvest the cash cow and buy back shares, because harvesting never really works.
Mag 7 era ending; broaden ownership.
AI is precipitating so much change that the Magnificent 7 framing is ending, exactly as FANG and FANG+ faded before it, and a new index of leaders will form that investors should be trying to identify now. Owning five to seven names that make up 80% of a portfolio is too much concentration risk; he keeps defaulting to roughly 25 companies, because in markets some things go right, some go wrong and you need enough positions to survive that. Part of why the Magnificent 7 became so large is structural: BlackRock and Vanguard have pushed everyone into indices, active managers have become closet indexers because one bad year ends them, and indexation forces everyone to be fully invested at all times rather than raising cash when multiples or conditions are poor.
Mag 7 correlation broke; rethink basket.
The Magnificent 7 was a set of seven highly correlated companies that sucked up all the attention and all the money and moved in unison dollar for dollar. That correlation has now broken down, which makes the relevant question who the real Mag X companies will be rather than continuing to own the seven. He endorses a roughly 25-name basket of future leaders, and notes that the optimal basket is partly public and partly private, so owning some random public company simply because it is public while ignoring names like SpaceX and Stripe would be stupid.
Wants humanoid and robotaxi leaders eventually.
Humanoids are a pretty exciting area and in his basket of about 25 future leaders he would want a humanoid company and a robot-taxi company. He thinks it is still a bit early: private valuations in these areas are inflated, with pre-revenue companies asking $30 to $40 billion, and his discipline is to wait until he can establish with roughly 75% confidence which company is the actual leader, accepting that he will have to pay more later rather than take a 1% chance on the wrong one.
Up Next
This All-In Podcast video, published May 09, 2025, features Philippe Laffont, Chamath Palihapitiya, David Friedberg, Jason Calacanis discussing VTI, Subprime consumer lenders, SMH, GOOGL, MAGS, Humanoid robotics, Autonomous driving. 10 trade ideas extracted by AI with direction and confidence scoring.
Speakers: Philippe Laffont, Chamath Palihapitiya, David Friedberg, Jason Calacanis · Tickers: VTI, Subprime consumer lenders, SMH, GOOGL, MAGS, Humanoid robotics, Autonomous driving