2025 Predictions: Tech, Business, Media, Politics!
Speakers
David Friedberg
— CEO, The Production Board
GavinSBaker
— Portfolio Manager, Atreides Management
Chamath Palihapitiya
— CEO, Social Capital
Jason Calacanis
— Angel Investor / Founder, LAUNCH
Summary
The All-In hosts Jason Calacanis, Chamath Palihapitiya and David Friedberg are joined by Gavin Baker of Atreides Management for the 2025 predictions episode, working through political and business winners and losers, the biggest deal, most contrarian beliefs, best and worst performing assets and most anticipated trends. The market-relevant calls cluster around AI and robotics - the year of the robot, Tesla FSD, a 2025 compute shortage, and high-bandwidth memory as the best performing asset - and around AI-driven software disruption, where enterprise application software and vertical SaaS are named the worst performing assets. Cross-asset views include Palihapitiya's long credit default swaps hedge against a mainline bank crisis and his warning on mega-cap index concentration, Friedberg's bullish calls on beaten-down Chinese tech and US nuclear buildout, and bearish views on legacy automakers, government service providers, legacy defense primes and US housing.
- Gavin Baker of Atreides Management joins the hosts for the annual bestie predictions show, with selected predictions to be posted as Polymarket markets.
- Friedberg calls 2025 the year of robots and autonomous hardware, citing Unitree's cheap quadruped and humanoid systems; Baker agrees and points to Tesla FSD going mainstream.
- Baker expects the world to run out of compute again in 2025, stays long Nvidia and chipmakers, and picks high-bandwidth memory makers SK hynix and Micron as the best performing asset.
- Palihapitiya's biggest winner is dollar-denominated stablecoins attacking the Visa and Mastercard duopoly; his biggest loser is mega-cap index concentration, and his best asset is long CDS protection against a small-probability bank crisis.
- Baker and Palihapitiya both name enterprise application software as 2025's worst performing asset as AI agents and cheaper agentic competitors undercut incumbents; Friedberg repeats vertical SaaS.
- Legacy autos, government service providers, IT services and old defense primes are widely seen as losers, with Tesla, Palantir and new defense technology as beneficiaries.
- Friedberg is bullish beaten-down Chinese tech stocks and a US nuclear power buildout; Baker keeps a no-China guideline dating to the Longtop fraud.
- Calacanis takes the other side on the Mag 7, likes Tesla and Google, sees autonomy-driven deals around Uber, DoorDash and Waymo, and is bearish OpenAI's valuation, MicroStrategy's bitcoin premium, legacy carmakers and US housing.
Ideas
2025 is the year of robots.
Friedberg predicts 2025 is the year of autonomous hardware and robotics: 2024 was the year of compute buildout and AI software rollout, and now capable robots are cheap enough to deploy at scale. He points to Unitree, whose Go2 quadruped costs about $1,600 with an API, lidar and onboard guidance systems and whose G1 humanoid costs roughly the price of a used car, and says his own team just ordered one for farm data collection. He expects humanoid robots and autonomous systems to arrive far faster than people expect - slowly, then all at once.
Tesla FSD hits mainstream adoption.
Baker profoundly agrees that 2025 is the year of the robots and says the clearest expression today is Tesla FSD: it already works, and it is going to cross into mainstream adoption and keep compounding at an accelerating rate. He notes he now actively prefers to ride in a Tesla running FSD when taking a late-night Uber because a tired human driver feels less safe.
Compute shortage keeps chipmakers bid.
Baker argues that the combination of reasoning and test-time compute demonstrated by OpenAI's o3 creates enormous new inference demand, so the world runs out of GPUs, accelerators and compute in 2025 exactly as it did in 2023 because new uses keep being found. Asked directly whether that makes him long Nvidia and the chipmakers, he answers absolutely.
Stablecoins scale and attack card duopoly.
Palihapitiya's biggest business winner for 2025 is dollar-denominated stablecoins. Two things happened in 2024: stablecoin usage decoupled from crypto volatility and started being used for wholesale business functions, and volume crossed a point of no return - roughly 1.1 billion transactions and $8.5 trillion of value in Q2 2024, more than double Visa over the same period. He expects an enormous number of use cases to move onto stablecoin rails, a political attack on high credit card costs at a time when consumer credit is already cracking, and the duopoly of Visa and Mastercard to finally be attacked. He thinks stablecoin usage could quadruple or quintuple by the end of 2025, and that taking 300 basis points of drag out of the economy would be worth about a trillion dollars in the US alone.
Stablecoins scale and attack card duopoly.
Palihapitiya's biggest business winner for 2025 is dollar-denominated stablecoins. Two things happened in 2024: stablecoin usage decoupled from crypto volatility and started being used for wholesale business functions, and volume crossed a point of no return - roughly 1.1 billion transactions and $8.5 trillion of value in Q2 2024, more than double Visa over the same period. He expects an enormous number of use cases to move onto stablecoin rails, a political attack on high credit card costs at a time when consumer credit is already cracking, and the duopoly of Visa and Mastercard to finally be attacked. He thinks stablecoin usage could quadruple or quintuple by the end of 2025, and that taking 300 basis points of drag out of the economy would be worth about a trillion dollars in the US alone.
Tesla and Google win AI 2025.
Calacanis picks Tesla and Google as the biggest business winners of 2025 because this is the year AI and robotics pay off for the companies investing most heavily in them. He is amazed by Google's AI comeback and singles out Gemini Deep Research as a step-function product that blows everything else in the market away, generating in minutes reports that consultancies would have charged a fortune for, and which will only get better given Google's access to search, YouTube, Gmail and Drive data. On the other side he is in awe of Elon's execution, citing Colossus and 100,000 GPUs stood up in about 45 days across Tesla and xAI, and expects this cohort to ship things people are not anticipating.
Avoid firms dependent on government revenue.
Baker's biggest business loser for 2025 is government service providers: in the age of DOGE you do not want the United States government at any level to be more than about 35% of your revenue, because for the first time the buyer is going to actually check the bill rather than spend wildly.
Index concentration sets up mega-cap drawdown.
Palihapitiya expects the absolute dollar drawdown of the mega-cap complex to be measured in trillions in 2025. His reason is not the fundamentals - he calls these exceptional businesses - but concentration: the top seven to ten companies are approaching 40% of the indices, and historic concentrations like that have generally foreshadowed a big drawdown that investors cannot inoculate themselves against. Even a 10% move is a couple of trillion dollars. He later proposes a prediction market on Mag 7 representation in the S&P 500 shrinking below 30%, implying dispersion into the rest of the index.
Legacy defense primes lose to tech.
Friedberg's biggest business loser is the old defense and aerospace providers such as Boeing and Lockheed Martin - defense 1.0 built around China dominance, large US defense budgets and cost-plus contracting. He expects budgets to shift toward a more technology and drone-oriented rationalization of pricing and spend, and he sees repeated failure at scale in these businesses, from Boeing's 2024 space program failure to problems in its airplane business, as they have become too clunky and bureaucratic. The new entrants - Palantir and the privately held Anduril - are the beneficiaries.
Legacy defense primes lose to tech.
Friedberg's biggest business loser is the old defense and aerospace providers such as Boeing and Lockheed Martin - defense 1.0 built around China dominance, large US defense budgets and cost-plus contracting. He expects budgets to shift toward a more technology and drone-oriented rationalization of pricing and spend, and he sees repeated failure at scale in these businesses, from Boeing's 2024 space program failure to problems in its airplane business, as they have become too clunky and bureaucratic. The new entrants - Palantir and the privately held Anduril - are the beneficiaries.
AI scrutiny crushes labour-arbitrage consultants.
Extending the government-contractor call, Palihapitiya says the traditional IT services and consulting firms - Wipro, Tata, HCL, Accenture, Cognizant and the large audit and consulting groups - are in real trouble. Cost-plus is a fancier way of saying time and materials, and time and materials is a fancy way of saying human labour arbitrage. With software entrepreneurs now advising the federal government on operations, and with AI making customers themselves more efficient, that spend gets extremely scrutinized in 2025 and these business models come under severe stress.
MicroStrategy's bitcoin premium is unsustainable.
On his shortlist for biggest business loser, Calacanis says MicroStrategy trading at two, three or four times the book value of the bitcoin it holds makes no sense to him and is not sustainable.
Truth Social's valuation defies its revenue.
Also on his loser shortlist, Calacanis flags Truth Social's listed parent as an obvious valuation disconnect: roughly $4 million to $5 million of revenue against a $7 billion valuation makes absolutely no logical sense to him.
Legacy automakers consolidate as Tesla wins.
Palihapitiya predicts 2025 brings the collapse of the traditional auto OEMs and a wave of mega-mergers, with the Honda-Nissan agreement at the end of 2024 as the signal for what the industry has to do. Tesla's vehicle quality, software quality and FSD autonomy are so far ahead that after a couple more meaningful product releases the public capital markets will simply conclude the legacy OEMs are uninvestable. He singles out the European makers as melting icebergs with no clear answer - it is not clear what Volkswagen does, or what Stellantis does - and expects pressure from smart investors to force them to merge.
Legacy automakers consolidate as Tesla wins.
Palihapitiya predicts 2025 brings the collapse of the traditional auto OEMs and a wave of mega-mergers, with the Honda-Nissan agreement at the end of 2024 as the signal for what the industry has to do. Tesla's vehicle quality, software quality and FSD autonomy are so far ahead that after a couple more meaningful product releases the public capital markets will simply conclude the legacy OEMs are uninvestable. He singles out the European makers as melting icebergs with no clear answer - it is not clear what Volkswagen does, or what Stellantis does - and expects pressure from smart investors to force them to merge.
Western automakers squeezed by Tesla, China.
Declining to discuss specific positions, Baker says he agrees 100% and adds his own reasoning: the legacy Western automakers are going to lose their Chinese business because they no longer make competitive products, and absent massive protectionism they end up caught between Tesla and the Chinese OEMs. The only real risk to the bear case is government intervention, since these companies are such large employers and are seen as national champions - but absent significant government support they are all in deep trouble.
Something big happens with Intel.
Baker expects a tidal wave of M&A in 2025 after four years in which nothing could get done, because there is an enormous amount of pent-up demand. His specific named case is Intel: something will happen with Intel and it will be big, and hopefully it is good for America.
Autonomy expands rideshare TAM; own Uber.
Calacanis argues the age of autonomy is here and that with the Lina Khan FTC era over, mega-deals and partnerships between Amazon, DoorDash, Uber, Tesla and Waymo can finally go through - Tesla could buy Uber for about 10% of its market cap, Amazon could buy DoorDash easily, and whoever combines autonomy with delivery, food delivery and e-commerce wins a ginormous space. He says he has significant exposure across these companies because ride sharing is still only about 1.5% of US rides and under 1% globally while the TAM goes to 20% in a short period, and Uber is his favourite because it has deals with eight autonomy partners and a global market.
Waymo takes share; big Waymo deal.
A specific 2025 deal candidate: Waymo. It launched in San Francisco in August 2023 when Uber and Lyft had 66% and 34% share; fifteen months later Waymo has 22% of SF rides, the same as Lyft, with Uber down to 55% - roughly 12 points of combined share taken in 15 months. It is now launching in LA, Austin and beyond, already runs in Phoenix, and has moved to a new hardware platform that should bring launch capex down significantly and improve ROIC, so the system works and is scaling. That sets up a massive financing, an IPO, or a merger or acquisition with one of the big ride-sharing companies this year.
Debt burden could crack a major bank.
Palihapitiya's most contrarian belief is a banking crisis at one of the major mainline banks. Adding up the total indebtedness of Pax Americana - government plus corporate plus mortgage debt, roughly $70 trillion - and sensitizing it to about 5% rates shows that in dollar terms today's rates are equivalent to 10% rates 25 to 30 years ago, when the debt stock was a fraction of the size. When the system collectively has to find three or four trillion dollars, a mark-to-market problem or a credit default problem among corporates or consumers can trigger a reserve issue at a mainline bank. He says he has two candidates in mind but will not name them, and stresses the probability is small, which is exactly why it is contrarian.
High-bandwidth memory makers outperform in 2025.
Baker's best performing asset of 2025 is the companies that make high-bandwidth memory. HBM is a bigger part of the COGS of a GPU than the TSMC content is, and today only two companies can really make it - SK hynix and Micron - with Samsung an open question until it gets its act together. HBM sits in Nvidia GPUs, AMD GPUs and Amazon's accelerators, it has been sold out for the last two years, and in a world where test-time compute and inference dominate it is arguably more important than it has ever been.
High-bandwidth memory makers outperform in 2025.
Baker's best performing asset of 2025 is the companies that make high-bandwidth memory. HBM is a bigger part of the COGS of a GPU than the TSMC content is, and today only two companies can really make it - SK hynix and Micron - with Samsung an open question until it gets its act together. HBM sits in Nvidia GPUs, AMD GPUs and Amazon's accelerators, it has been sold out for the last two years, and in a world where test-time compute and inference dominate it is arguably more important than it has ever been.
Buy cheap CDS protection for asymmetry.
Palihapitiya's best performing asset pick is to be long credit default swaps - buying protection that there is no default event in 2025. He frames it explicitly as insurance with extreme asymmetry: 92 times out of 100 it goes to absolute zero, six of the remaining eight times you make about 10x, and twice you make 100x to 1000x. Given the concentration of the S&P, the total gross amount of debt outstanding and the risk of rates spiking, a little insurance may not be a bad thing, and if it hits it would be the equivalent of Ackman's CDS trade at the start of COVID. Access requires an ISDA and pricing from the big investment banks.
Cheap Chinese tech rerates on deal.
Friedberg's best performing asset is Chinese tech stocks or Chinese tech ETFs, which everyone has dumped over the last couple of years on an isolationist positioning stance. He sees three drivers: the Trump administration lining up a grand bargain with China, signalled by the request to halt the TikTok ban, that opens the Chinese market to American companies and vice versa; an unfathomable Chinese electricity buildout including a newly approved $137 billion hydroelectric project plus nuclear, which lowers cost per kilowatt-hour; and the Chinese Communist Party's ability to throttle free markets and entrepreneurship back up to motivate a deal. These stocks look cheap - he notes Alibaba trades at a decent multiple - and Chinese firms with the best unit economics and cost of production are positioned to serve global demand for automation and rebuilt manufacturing capacity.
Avoid Chinese equities despite cheap multiples.
Baker takes the other side on China as a matter of policy: he has kept a no-China guideline ever since the Longtop financial fraud more than 15 years ago, where great investors owned a name with a Western auditor while the documents were being swapped at the local post office. He thinks it is simply a hard place to make money if you are not Chinese. He concedes there are really high quality Chinese companies at mid-single-digit multiples serving a global market, and that if Trump and Xi do the deal that leaves Putin out in the cold the upside would be large.
Mega-caps compound earnings via internal AI.
Calacanis explicitly takes the other side of Palihapitiya's concentration call and picks the Mag 7 as the best performing asset of 2025. His reason is margin expansion: these companies have learned in the last couple of years how much earnings growth they can create by not hiring, outsourcing roles and automating, and because they build AI for other companies and consumers they apply it internally first. That internal application produces earnings growth over the next couple of years that people will not be able to comprehend.
Agents crush enterprise application software.
Baker's worst performing asset for 2025 is enterprise application software. He expects 2025, particularly the second half, to be the year of agents - AI models that take action on your behalf and can do anything online a human can do. If the labs and the big cloud providers dominate agents, which seems likely given that the lowest-cost producer wins, enterprise application vendors are in a lot of pain: they talk a big game about agents but do not own their own models or their own compute, and their customer data moat and relationships are weak because most customers also have relationships with AWS, Google or Microsoft.
Agentic AI undercuts legacy enterprise software.
Palihapitiya doubles down on the same call with his own framing: the software industrial complex. These large, bloated enterprise software companies have convinced enormous numbers of organisations to pay for heuristics and business rules wrapped around a CRUD database, and what they have really perfected is a go-to-market motion of golf trips and steak dinners that equates to no product value. In a world of agentic software these workflows can be rebuilt far more efficiently - his own 8090 got to 80% of the traction in under a year with a tiny engineering team - so a new entrant can answer a $100 RFP with $10 and still be hugely profitable, while CEOs and CFOs pressure CIOs on spend. Fissures appear in those businesses in 2025.
Per-seat vertical SaaS pricing gets compressed.
Friedberg triple-underlines vertical SaaS as a worst performing asset for 2025: the per-seat pricing model is being challenged and pricing compressed as companies explore in-house tools built with AI that replace these traditional business applications.
Stretched consumers hurt legacy car makers.
Calacanis reasons that consumers make five big decisions - college, spouse, kids, cars and homes - and only two of those are tradeable, while the consumer is up against record debt. On cars, the legacy OEMs have overbuilt and there are tons of cars sitting on lots at a moment when the consumer cannot afford them, so legacy car companies face continued headwinds and are terrible assets in 2025.
Affordability and supply hit US housing.
The other half of Calacanis' worst performing asset call is US residential real estate: people cannot afford homes at current mortgage rates, and in places that allowed building - he cites Texas - housing values and rents have now fallen two years in a row, with the same pattern appearing in other states that permit supply while people leave the states that do not.
US nuclear buildout accelerates in 2025.
Friedberg's most anticipated trend is the announcement of nuclear power buildout in the United States in 2025 as a function of deregulation and new technology. He argues the US must add more electricity production capacity than any other renewable source can scale to in order to meet competitive demand against China, so nuclear is an inevitability; he expects the deregulation to happen in 2025 and cites as a leading indicator the number of very smart people leaving good jobs to start nuclear power companies in anticipation. He says he is very bullish.
Up Next
This All-In Podcast video, published January 04, 2025, features David Friedberg, GavinSBaker, Chamath Palihapitiya, Jason Calacanis discussing ROBO, TSLA, NVDA, SMH, STABLECOINS, V, MA, GOOG, Government services contractors, MAGS, BA, LMT, PLTR, IT Services sector, MSTR, DJT, VOLKSWAGEN, STLA, European automakers, INTC, UBER, DASH, LYFT, KBE, 000660.KS, MU, 005930.KS, Credit default swaps, KWEB, BABA, FXI, Enterprise Software sector, Legacy enterprise IT, Vertical SaaS, Legacy automakers, US residential real estate, URA. 32 trade ideas extracted by AI with direction and confidence scoring.
Speakers: David Friedberg, GavinSBaker, Chamath Palihapitiya, Jason Calacanis · Tickers: ROBO, TSLA, NVDA, SMH, STABLECOINS, V, MA, GOOG, Government services contractors, MAGS, BA, LMT, PLTR, IT Services sector, MSTR, DJT, VOLKSWAGEN, STLA, European automakers, INTC, UBER, DASH, LYFT, KBE, 000660.KS, MU, 005930.KS, Credit default swaps, KWEB, BABA, FXI, Enterprise Software sector, Legacy enterprise IT, Vertical SaaS, Legacy automakers, US residential real estate, URA