The AI Selloff Doesn't Match the Data | Top AI Investor Explains
Video Overview & Insights
In this episode of Invest Like the Best, Patrick O'Shaughnessy sits down with Gavin Baker to dissect the intense AI and semiconductor market sell-off of July 2026. Describing the period as "2022 packed into a single month," Gavin explains why public market panic contrasts sharply with the booming fundamentals seen on the ground in Silicon Valley. They dive into the accelerating operating cash flows of hyperscalers, the soaring spot prices for GPUs, and why open-source AI models are actually driving massive infrastructure demand. Gavin also unpacks the game theory behind Long-Term Agreements (LTAs) in the memory market, NVIDIA's strategic dominance, SpaceX's massive data center ambitions, and the looming regulatory risks for AI compute.
Interesting conversation but I would like to make two points -
1) Demand growth is not equal to exceptionally high valuations (NVDA & TSM are not that expensive but everything else is)
2) What's good for silicon valley founders and VC is NOT automatically good for "the world", in fact most of the time rest of us pay the price
Timestamps:
0:00 Intro
If you have a machine that turns $1 into $5 (data center). Why wouldn’t these companies buy them 100% on their own?
1:17 AI Selloff vs. Fundamentals
10:20 Financing the AI Buildout
You both need a good dose of Ed Zitron...everything you say here is nonsense...
18:18 GPU Prices Keep Rising
24:23 Claude Moves Markets
First ten minutes comes off as absolute cope 😅
29:01 What Could Break the Thesis
36:59 The Memory Supply War
42:08 Nvidia’s New Playbook
50:14 China and Open Source
61:51 Data Centers and Regulation
71:11 SpaceX and Orbital Compute
Key insights by Grok:
• July 2026 AI/semiconductor selloff was “2022 in a month” (AI names down 40-60%), but fundamentals accelerated.
• No negative quantitative metrics found: GPU availability, rental prices, DRAM spot prices, and token growth all rose. Nvidia at its lowest forward PE in 10 years.
• Market lacks visibility into private players (Anthropic, OpenAI, open-source inference clouds). Open-source models (e.g., GLM 5.2, Kimi) shift share from high-margin frontier tokens but increase total compute demand equally—driving more revenue to infrastructure.
• Old GPUs still rising in price in 2026 (unexpected). Long-term contracts underprice current spot; as they roll off, compute reprices higher and hyperscaler operating cash flow accelerates (e.g., MSFT/Meta/AMZN from ~28 to 32-35).
• Meta renting compute and open-source news were misread as bearish; neither signals slowdown. Credit/real yields are a real risk but majority of buildout still funded by cash flow.
• Other topics: memory LTA game theory, Nvidia strategy, China/open-source, data-center regulation, SpaceX orbital compute ambitions.
#Investing #AI #Semiconductors #Nvidia #StockMarket #Technology #DataCenters #OpenSource #Finance #VentureCapital #nvidia #stocks #sell #stockmarket
Presented by Ramp:
i assume Gavin knows what he's talking about. but he's not a good communicator. quite difficult to follow what he's saying because he uses terms like "they", "you", "them", "us", etc, instead of just being specific about which entity he's talking about so it's easier to follow
https://ramp.com/invest
Sponsored by Vanta, WorkOS, Rogo, and Ridgeline:
The drinks are always a continuity giveaway. Gavin's green drink at 1:11:00 somehow gains four ounces in 24 seconds. 🤷
https://www.vanta.com/invest
https://workos.com/
How is he so sure they aren't building on dept ?
https://rogo.ai/invest
https://www.ridgelineapps.com/
This guy is so out of touch. The impacts of data centers on water, electricity costs, etc is just comical. Can’t tell if it’s stupidly or pure greed to push such un-levelled takes. Either way he’s drunk off the kool aid
******
Patrick O'Shaughnessy is the CEO of Positive Sum. All opinions expressed by Patrick and podcast guests are solely their own and do not reflect the opinion of Positive Sum. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. Clients of Positive Sum may maintain positions in the securities discussed in this podcast. To learn more, visit psum.vc
What I conclude from his talk is crwv is the best place to be for next couple years
More User Perspectives
ABSOLUTE CANNOT MISS EPISODE WHEN THIS GUEST IT ON
@yesindeed900This guy has serious sinus problems 😮
@arturosandovalThe final parabolic push now before the bubble bursts.
@ntf81"If you adjust for that, '28 - '35"; yeah, we'll see.
@kaczynski2333This guy’s tone of voice is giving “highly leveraged”
@cameronmoore6663I think the issue with the current Vendor Financing is it just takes 1 bad player to take the whole chain down.
@Snowy123the hardware companies will always be fine. it's openai, anthropic, and spacex that are doomed. we will all be getting our AI from the lowest bidder, not them.
@LuminairPrimeHard to like Gavin’s verbal communication. Can someone just summarize?
@Simon1234-e2xI’m about to delete all my trading apps. Spent the last 6 months studying charts, buying courses, and trying to learn the ropes, but I’m still just losing money. Every time I take a trade, the market immediately reverses on me. Being a beginner in this market feels impossible right now. Is anyone actually making consistent money or is this whole thing just a scam? 😭...
@RichardCastro-f4xHe sounds like he's about to cry the whole time... Similar to how Bill Ackman sounded in that Covid interview
@nutcracker4986Retired at 48, earning 45K biweekly... but honestly, the money isn't even the highlight. What means the most is having a family that stood by me when life was tough. I'm living the life I once prayed for. If you're still fighting your battles, don't lose hope your season is coming.❤️
@Barbara-i7i9qIm not gunna act like I am a AI guru and I know whats gunna happen but one thing i do know is the negative people in the comments even if they are right on AI will never truly prosper maybe in the short term if AI is a bust but in the long run optimist always beats the pessimist also willing to bet most have under a 6 figure portfolio lol
@gymislife1994Why does this guy sound like he's about to cry?😂 I'm smelling desperation from these tech guys, they haven't produced anything groundbreaking in 15 years and they're getting desperate. They're massively betting AI is the answer, but the demand to justify the spend won't be there.
@mateo8120you really should put a "recorded on" timestamp or comment somewhere (commenting on 8/4 after a rip in Q's and SOX)
@a0abesIf I never heard Gavin Baker talk again that would be fine with me.
@thoughtsandruminationsLiteral local market peak on the day this was dropped
@alexgrosek7649.80 speed helped me immensely
@BillyBoggleAs ever, investors their own worst enemies.
Being concerned about hyperscalers taking on debt is asinine - lenders should be chomping at the bit to extend them credit. Have people forgotten why theyre hyper in the first place? Theyre the most profitable entities in the history of the known universe, and thus, the most credit-worthy.
Besides, they already have the forward orders on their books to pay back those loans - theyre practically as safe, or more, as U.S. bonds.
Just off vibes...the manic energy looking for something wrong and not finding it means there are blind assumptions somewhere. Theres always a downside and not noticing until the last minute means.....
@Eggs-n-JakeyI like Gavin but all he does anymore is promote his holdings
@Christopher-q7y76his phlegmy voice literally makes me want to spew
@trentcrawley3502i like how they are not wearing headphones, many americans do that in streams, for no real reason at all
@stephenmeier7053Gavin is just looking at GPUs, he seems oblivious to the fact that NVDA GPUs are Ferraris and you don't need Ferraris to go to the grocery store everyday. Custom silicon from GOOGL, META, MSFT, AMZN can do most jobs and cheaper at that. Add in competition from the Chinese and you have an unsustainable 75% NVDA margin. MSFT will be using Kimi-K3 and the savings from that is $600 million! Plus, add the bogey man of when the hyperscalers says no mas on cap-ex and the AI Hardware trade compress big time. This is one of the main reason of the selloff, no one wanted to hang around and find out that one of them might say "no mas". Plus, big investors has already made a ton of money and so they closed their books in June with a massive profit and went shopping somewhere else. It is literally quite that simple Gavin.
@meindaknowBro fell asleep in a tanning bed before this podcast
@EloAngelLoLI believe that a 'hard economy' is actually the best time to build if you have discipline and the right strategy. Most people spend their time complaining but I believe in execution over excuses. I made a smart move in early 2026, and because I believe in staying the course, I just hit $190k. The momentum is still building, and by the end of this year, it's going to be massive.
@MayorJoachim01The sad truth is that no one has a clue, we all react to what happens as it happens and try to analyse it but can’t predict an iota of what is going to unfold in the markets… content creators are like amplifiers, when times are good they affirm it and try to tell you why it’s good and that it’s looking bullish but then all of a sudden the market turns bearish and everyone affirms it again and try to analyse why… it’s so sad that many are so powerless and it's not about guessing the market's next move; it's about playing it smart and steady during trading...managed to grow a nest egg of around 2.3Bitcoin to a decent 19Bitcoin in the space of a few months.Thanks once again, Marie j Freeman., for your consistent updates and continued support.
@abdelatiifarife1114Gavin said "you know" 74 times this interview (according to Gemini). I would have picked the over.
@ShurtliffChannelOne metric that isn’t accelerating. Profit from AI.
@AshtypantsLet’s talk about off balance sheet debt. Somewhere between $1-$2 trillion.
@AshtypantsI'm a beginner and want to start trading! how can I make high profits having just 5k to invest thank you 🙏
@KimaLok-o1jBEAR DOOM #1.
Credit crisis coming due to massive capex "burn"-
Two year contracts for hyperscalers come up soon and range of that re-pricing higher is near 100%. Even cut in half to 50% and this pays for all the capex out of OCF to 28'. Probs comes in somwehere in between but a re-pricing of at least 25% higher makes this a non-issue
BEAR DOOM #2-
Circular financing is running the show-
The bear lens here starts by looking at RPO/Backlogs for clouds---They are huge at $1.5T for top 3 clouds but these commits are out a decade. For AWS the total of frontier labs is max at peak is only 16% of ebitda for AWS and max 15% of revenue when one takes the whole AWS cloud biz. Zitron's analysis only considers "direct AI" revenue which is still in the 1st inning.
It is of course a higher % of "Direct A.I. revenue" but every $1 of AI revenue is pulling an additional $1.5-$2.5B of new AWS cloud revenue at 40% margins. So Ed Zitron's doom case on circular has merit at the surface but if ya know math and the whole picture it has little merit.
BEAR DOOM #3-
Open weight models will kill the frontier models, the hyperscalers and their circular financing.
AWS was designed for a multi-model token gateway world and MSFT Azure is catching up. Little known fact is they are "rooting" for more open wgt traction because
A- Their revenue retention is 95-100% vs. only 60% for frontier models(Amzn AWS pays ~40% partner fees to ANT/OAI)
B- Custom services and AWS attach, margin is much higher on open wgt models than frontier base IaaS compute
C- Open wgt models drive 2-3x volumes
D- They get more negotiating leverage vs. frontier models as the # open wgt. models rise and being the ONE place to optimise your mix of the right models for the right work and payoff
*They WANT a higher open wgt mix as the ROIC is higher but for sure don't want the frontier models to die. Maybe 65-35% mix of open wgt to frontier is optimal so both the frontier revenue commitments(circular) come through AND open wgts drive scale and Jevon's Paradox. But the idea that AWS would risk its entire biz by over-risking on frontier over dependence is not valid in the numbers nor the strategy.
Take a shot every time Gavin says Aham aHaM AhAm Aham Ahhhmmmmm
@UnsaltedCashew38Actually the posting of this again aligned with a new ATH.
@IndiaEditionTell me something negative - nvidia buys its owns stock, has greed to backstop its customers and open ai burns through cash faster than anything on earth…..These forward returns Gavin, are they in the room with us right now?😂
@RushmanismWe need an debate with Gavin Baker and Ed Zitron
@abekawser4908i love a good pump
@Testing12344dhThis guy should talk to Ed Zitron 😊
@eHex_MaxiHUGE Selloff!
@c4leon