andrey

australian in london. building. ex-goldman sachs M&A. prev world #1 concert guitarist. 2:47 marathoner.
side project

[andrey @andreylebedev29]

this guy cooks!! signal to noise through the roof

[amit @amitisinvesting]

A TON OF THINGS HAPPENED IN THE STOCK MARKET TODAY.

Here's a full recap:

1. The U.S. reportedly offered Iran a deal to halt the siege and lift sanctions in exchange for reopening the Strait of Hormuz and ending proxy attacks, according to Al Arabiya. Axios also reports that Rubio told several foreign counterparts the U.S. does not plan new strikes on Iran for now, with pressure shifting toward the naval blockade and new sanctions campaign instead. Crude Oil fell 4% and the 10-year treasury bond fell from 4.72% to 4.62%.

2. Global physical gold-backed ETFs $GLD attracted $6.4B of inflows last week, their largest weekly intake since January and the 3rd-largest weekly inflow on record. North America led with $4.4B, followed by Europe at $1.7B and Asia at $300M. This marked the 7th straight week of inflows, with global gold ETFs pulling in $16.4B over that stretch. Total AUM in global gold ETFs rose by $33B last week to $615B, the highest level since the second week of May.

3. Intuit $INTU reported Q4’26 revenue of $4.4B, beating estimates of $4.27B and up 14% YoY. Adjusted EPS came in at $4.03 versus $3.58 expected. Global Business Solutions revenue rose 14% YoY to $3.4B, the Online Ecosystem grew 17% YoY to $2.6B, Consumer revenue increased 14% YoY to $930M, and Credit Karma revenue rose 16% YoY to $743M. For FY27, Intuit guided revenue to $23.3B–$23.5B versus $23.72B expected, while adjusted EPS guidance of $22.88–$23.12 came in well below the $27.32 estimate. The company also raised its dividend 15% YoY to $1.38/share, bought back $5.5B of stock, and has $7.9B remaining on its authorization. Management said its strategy is to win as an AI-driven expert platform while staying disciplined on investments and scaling its big bets.

4. President Trump said the U.S. Navy has removed and/or detonated all mines from international waters in the Strait of Hormuz. He said Iran has been notified that any ship or boat placing new mines will be “immediately and systematically destroyed.” Trump added that Space Force is monitoring every square inch of the Strait, along with Pickaxe Mountain and the three previously destroyed nuclear sites, and said a “Zero Tolerance” policy on mine placement is now in full effect.

5. Canada is responding to U.S. tariffs with new tariffs of its own. The country is raising steel tariffs to 50% from 25%, while roughly 700 products will face new tariff rates of 15%, 25%, and 50%. The measures are set to take effect on September 8, marking another escalation in the U.S.–Canada trade dispute.

6. Anthropic is expected to tell IPO investors its total addressable market exceeds $30T, topping SpaceX’s $28.5T estimate, according to WSJ. The figure represents the potential value of work Anthropic believes AI models could eventually perform, not a direct revenue forecast. Anthropic generated $11.6B in Q2 revenue and could seek to raise as much as $100B at roughly a $2T valuation. IPO documents are expected within weeks, potentially setting up a September or early October listing.

7. OpenAI’s data-center head Chris Malone left the company last week, according to WSJ. Malone joined in March 2025 shortly after Stargate was announced and played a key role overseeing OpenAI’s massive data-center buildout. He previously led data-center strategy at Meta and earlier worked on data-center technology at Google. The departure comes just weeks after OpenAI also replaced its chief revenue officer, adding another senior leadership change as the company races to scale infrastructure, revenue, and compute capacity.

8. ClickHouse has surpassed $350M in annual recurring revenue, up 40% since May, as AI agents drive demand for database and observability infrastructure. OpenAI’s usage has reportedly grown roughly 10x over the past year to more than 30 petabytes of data per day, or around 30T events daily. OpenAI has also shifted parts of its log-management workload from Datadog to ClickHouse over the past year. ClickHouse was valued at $15B in January and says gross margins currently range from 50%–70%. Earlier this year, the company acquired Langfuse to expand deeper into monitoring AI applications and agents. Nebius $NBIS owned a 28% stake in ClickHouse as of May 2025, though that stake has likely been diluted by subsequent fundraising.

9. JPMorgan reiterated its Overweight rating on SpaceX $SPCX with a $240 price target, saying the company’s AI ambitions are coming into sharper focus and that it is increasingly positive on Grok. The firm highlighted SpaceX’s completed acquisition of Cursor on 8/14 as an important step in building enterprise AI capabilities. Cursor brings roughly $4B of ARR as of June 2026, with about 75% coming from businesses, which JPMorgan says should help streamline go-to-market and provide valuable model-training data. The firm also said Cursor data is already showing up in Grok’s supplemental training, with tangible improvements in recent model performance.

10. OpenAI says its new Broadcom-built Jalapeno AI chip outperformed Nvidia $NVDA GB300 in both throughput per watt and response latency during internal testing, according to Bloomberg. The chip is built specifically for inference, not training, and runs at roughly 700 watts. OpenAI plans to begin deploying Jalapeno for its models later this year, saying the performance gap widened on larger workloads, including Moonshot’s Kimi model, and that the chip has also performed well on unreleased OpenAI models. The key caveat is that Jalapeno was tested against GB300, not Nvidia’s newer Vera Rubin generation. OpenAI says a second-generation chip is already nearing tape-out, while work on a third generation has begun.

11. The top 10 most active options today by contracts traded were $NVDA with 1.8M contracts, $TSLA with 1.8M contracts, $AAPL with 636K contracts, $SPCX with 548K contracts, $INTC with 540K contracts, $AMZN with 498K contracts, $MU with 483K contracts, $AMD with 403K contracts, $PLTR with 361K contracts, and $SOFI with 359K contracts.

12. Raymond James raised its Nvidia $NVDA price target to $352 from $330 and reiterated a Strong Buy rating. The firm says Nvidia’s CPU opportunity is becoming more important, especially for agentic AI workloads, even though CPUs are only about 3% of sales today. Raymond James expects CPU revenue to reach roughly 5% of total revenue by CY28 and believes Nvidia could potentially become the world leader in CPU revenue within several years. The firm also argued the stock remains inexpensive, trading at less than 15x CY27 GAAP earnings, below the S&P 500 at 18.6x, despite sales and net income growth still expected to exceed 20% in CY28. Its new $352 target is based on a 22x multiple on CY28 estimates, which Raymond James views as conservative given Nvidia’s leadership, CUDA moat, GPU performance, free cash flow, and history of trading at much higher multiples.

WALL STREET IS THE GREATEST SHOW ON EARTH.

[andrey @andreylebedev29]

@gearside please no

[andrey @andreylebedev29]

@hannahhaina i cannot read these ai slop pieces. learn to write yourself, please

[andrey @andreylebedev29]

@patrick_oshag so cool this is being recognised!

[JUNK BOND ANALYST @junkbondanalyst]

Josh Kushner wrote his first-ever formal letter to backers of his $65 billion investment firm, Thrive. The letter is an opportunity to understand what Thrive is seeing real-time, and to better understand how Thrive is applying its strategy to this current moment.

Letter to Thrive Capital Investors:

In Thrive’s history, I have never written a formal investor letter. This year, because we will not be together for an Investor Day, I wanted to write to you directly. We are deeply grateful for your trust, and we feel a profound responsibility to continue earning it through our work.

The principles that have guided Thrive since inception remain the same. What has changed is the world around us. I want to use this letter to share what we are seeing, how we are applying our strategy to this moment, and why the same principles that brought Thrive here continue to shape the work ahead.

Our Strategy

Thrive began with a simple but unconventional idea. We believed an investment firm could be opportunistic across stage, sector, and geography, while remaining deeply concentrated in a small number of people and ideas. We wanted to build a firm with the freedom to follow its curiosity wherever the best opportunities emerged, and the discipline to act only when that curiosity became conviction.

We have never believed that the traditional boundaries of our industry needed to define us, and we remain grateful to the partners who believed in this approach long before it was obvious. Our strategy has never been to be different for its own sake. Rather, the strategy has been, and always will be, to think from first principles about what founders need, where the best opportunities are emerging, and how we can build Thrive to concentrate our time, capital, and energy on the people and ideas we believe in most.

We have long believed that a small number of exceptional companies create a disproportionate amount of value and can compound their advantages for far longer than the market expects. We also believe that being a meaningful partner requires time, context, and trust. You cannot know a company deeply from a distance. You cannot be a true partner in difficult moments without the time and context that concentration requires. That belief shaped the portfolio. It also shaped the firm.

Since our founding, the scale of Thrive has changed, but the strategy and ethos of the firm have not. In every way, scale raises the standard. Today, Thrive manages more than $65 billion, more than half of which is driven by investment gains. But AUM will never be the product. We view it as a reflection of trust earned over a long period of time. The product of Thrive is the quality of our judgment, the depth of our partnership, the discipline of our process, and the culture of the team making decisions every day.

Culture is a word that is easy to overuse and hard to define. For us, it has always meant a small number of simple things taken seriously.

We try to listen more than we talk. We try to say “I don’t know” when we do not know. We try to be intellectually honest even when the truth is uncomfortable. We try to be ambitious while remaining self-aware. We try to be kind without being weak. We try to be tough without becoming callous. Gratitude without ambition can become nostalgia. Ambition without gratitude can become entitlement. We need both. We try to focus on our craft rather than the noise around us.

This requires humility. It requires recognizing that the most important insight may come from the least experienced person in the room. It is not possible to always make the right decision, but it is possible for us to always have the right conversations. Above all, we understand that we will never be a finished product, either as a firm or as individuals. Excellence is not something we arrive at. It is the work itself.

While I am proud of how far Thrive has come, and thankful for all we have built together, I am even more aware of how far we have to go. That tension has always been important to us as it has been part of the ethos of the firm from day one. We never confuse progress with arrival; as soon as we summit a mountain, we immediately look for the next climb.

We remain early in our journey and a small percentage of our full potential. That reality is both humbling and motivating.

Our Perspective

The clarity of our thinking matters more today because the opportunity before us is so significant. There are moments in technology when progress feels linear, and moments when it begins to compound in ways that are difficult to understand in real time. Today feels like one of those exponential moments. Ideas that would have represented major breakthroughs only a short time ago now appear almost weekly. The pace of technological and scientific progress requires us to keep updating our understanding of what is possible, while remaining disciplined about where enduring value will accrue.

Artificial intelligence is the most important technology paradigm of our lifetimes.

The internet democratized access to information. AI has the potential to democratize access to intelligence. If that happens, the implications could be extraordinary. Every person, company, and institution may eventually have access to capabilities that were previously unavailable or unimaginable. Workflows will change. Products will change. Cost structures will change. Entire industries will be rebuilt, and the societal implications of both the magnitude and rate of change are likely to be profound.

Our work with OpenAI has given us a close view of how quickly these models are becoming useful in the real world. The most important lesson is not any single benchmark or release. It is the rate of improvement. Systems are beginning to operate with greater autonomy, and the amount of work a model can complete in one uninterrupted effort is expanding quickly. For us, the conclusion is not simply that AI will create value. It is that we must be increasingly precise about where value will be captured, and where it will compound over time.

One area we believe durable value will be created is in the infrastructure required for AI to scale: compute, power, networking, memory, advanced manufacturing, and the supply chains that enable models to be deployed at scale. As AI adoption grows, these inputs become increasingly important, and in some cases, increasingly scarce. These constraints will help determine the pace at which the shift occurs.

It is difficult to overstate the magnitude of the opportunity. It would also be a grave error in our minds to let excitement weaken our investment discipline. In moments of euphoria, investors tend to convince themselves that second- and third-tier assets are actually first-tier. The story becomes so powerful that people stop making distinctions. Capital flows toward the theme rather than the company. Markets begin to reward proximity to the future rather than evidence of enduring quality. Within Silicon Valley in particular, the industry can become fixated on hyper-incremental technological turns rather than where the technology ultimately leads.

We have seen this before, and we will see it again. Technological optimism and investment discipline are not opposites. In fact, the more optimistic we are about the size of a paradigm shift, the more disciplined we must be about where value will be captured. Not every company exposed to an important theme becomes important. Not every company using new technology becomes enduring. Not every fast-growing business is exceptional. And not every exceptional company is a great investment at every price. Our responsibility is to maintain those distinctions.

We are independent because markets move between fear and enthusiasm, and neither is a substitute for judgment. In difficult moments, independence can mean moving toward an exceptional company while others move away. In euphoric moments, it can mean remaining patient while activity around us accelerates, knowing there will be future opportunities to partner with many of these same exceptional businesses but with a more attractive risk/reward calculus. It is only because of repeated lessons throughout Thrive's history that we have learned the highest-quality assets compound over very long arcs and, as such, patience, discipline, and conviction must sit together in equal measure.

As we look at the world today, we increasingly organize our opportunity set across four categories:

Emerging Technologies: These are businesses creating the building blocks of both new industries and new paradigms which will drive exponential change. Companies such as OpenAI, SpaceX, and Isomorphic represent what we believe are among the most important technology platforms being built anywhere in the world. They are difficult, deeply ambitious, capital-intensive, and, early on, very often misunderstood. We do not view these as reasons to avoid these businesses, and in many cases, they are part of what makes them important and defensible.

Infrastructure and Applications of Emerging Technologies: These are businesses such as Stripe, Databricks, Anduril, and others whose products, data, distribution, and customer relationships can position them to be more valuable in an AI-first world. Some will underpin the infrastructure layer. Some will create domain-specific intelligence. Some will apply AI to existing workflows in ways that fundamentally improve the customer experience. The common thread is that they are deeply advantaged by technology paradigm shifts.

Traditional Industries to be Transformed by Emerging Technologies: This is the work of Thrive Holdings.

Since seeing the preview of ChatGPT in 2022, we have believed that AI would transform large, legacy industries that collectively represent several trillion dollars of enterprise value. But unlike many of our peers, our conviction was not only that these industries would be disrupted from the outside in but also that many would be transformed from the inside out.

The most important ingredients for applied AI are often proprietary data, domain expertise, workflow ownership, and the ability to implement change in real-world operating environments. Existing businesses often have these ingredients, but they do not always have the technology, talent, urgency, or ownership structure required to leverage them fully. That is why we started Thrive Holdings.

Through Holdings, we have acquired more than 70 businesses that serve tens of thousands of customers. We now have a dedicated team of 35 engineers and operators working alongside the operators inside these businesses. On our accounting platform, we have built custom agents that have achieved 98% accuracy and reduced preparation times by up to 30% for tax returns. At our IT services firm, triage and resolution agents have reduced ticket completion time by 60% and are resolving 50% of all tickets end-to-end.

We are proud of this scale and early impact. What is most exciting at Holdings is the progress happening inside the businesses themselves. In partnership with OpenAI, we are building increasingly capable custom models and proprietary systems that can change how these companies operate

This work has also taught us that AI transformation is not only a model-layer problem. Models matter enormously. But real-world transformation requires product thinking, engineering, workflow design, incentives, training, and organizational change. It requires people who understand the business and people who understand the technology working together in a very deep way. In many ways, change management is the product. That is hard. It is also why the opportunity exists.
Assets Counterpositioned to Transformation: This is the work of Thrive Eternal.

Most of our time is spent thinking about the first-order consequences of AI. What becomes more efficient? What becomes cheaper? What becomes automated? What new products become possible? But there will also be second- and third-order consequences. In a world of abundant intelligence, certain scarce human experiences may matter more. In a world of fragmented distribution, trusted institutions may matter more. In a world where content and software become easier to create, assets rooted in identity, culture, community, history, and physical experience may become more valuable.

Eternal was created to invest in a small number of these assets and steward them over very long periods of time. The objective is not to buy, optimize, and sell. This distinction is very important. We intentionally chose a permanent capital structure because the assets we want to own deserve a time horizon measured in decades.
Each of these categories is distinct, yet the philosophy behind them is the same, and the strategic connectivity among them is foundational to the firm's strategy.

The frontier companies help us understand where technology is going before it becomes obvious. That perspective makes us better investors in both the public and private markets, where we can identify the companies best positioned to benefit from new technology paradigms long before those advantages are fully reflected in their businesses or valuations. It also makes us better partners to the exceptional founders building the infrastructure and applications that bring these technologies into the real economy.

Through Thrive Holdings, we learn what it actually takes to transform established businesses from the inside out, giving us hard-earned operating insights. And through Eternal, we better understand the characteristics that enable businesses to retain their relevance and compound through periods of significant change. That perspective strengthens our judgment around durability and long-term value creation across the firm.

At the center of all of this is a simple belief: our job is to partner with the most talented and ambitious entrepreneurs in the world, develop a deep understanding of both technology and its practical implications, and build an organization where every lesson compounds across every strategy. Over time, we believe that this shared knowledge and perspective will become one of Thrive's most enduring competitive advantages.

Our Work

At the early stage, we remain focused on maintaining close contact with the frontier. Early-stage investing is the largest universe of opportunity and the clearest way to understand how the next generation of founders is thinking. We have been fortunate to partner with companies such as Cursor, OpenEvidence, and Physical Intelligence, which are helping define progress at the model, application, and infrastructure layers. We expect to continue investing at our historical pace through the remainder of the year, with a focus on emerging technologies, AI applications, and infrastructure.
At the growth stage, we have built positions in existing growth-stage portfolio companies we have admired for a long time, including Anduril, Isomorphic Labs, Stripe, and OpenAI. We are humbled that founders increasingly seek Thrive as a lead partner. We will never take that for granted. Access is earned through judgment, work, and the quality of partnership after the investment is made. When Thrive leads an investment, we want that decision to communicate a deep belief in the quality of the team, the importance of the product, and the magnitude of the opportunity.

We are also deepening our work in public markets. Historically, Thrive has made a small number of public market investments when we believed a business was meaningfully misunderstood. Today, we believe that understanding the most important public technology companies is increasingly essential to our work across the firm. Many of our growth-stage companies compete with, partner with, or become the next generation of public technology companies. The more precise our understanding of public markets, the better our judgment will be in private markets. Our targeted public market work is an extension of our desire to continuously learn, to sharpen our collective judgment, and to improve our ability to evaluate quality irrespective of where it exists.

Our Performance

We know that you have entrusted us with capital that supports universities, hospitals, foundations, families, countries and institutions whose work extends far beyond our own. We never forget who we work for. When we do well, the causes and people you support do well too.

We are ultimately measured by the returns that we generate. That is appropriate. But returns are outputs. If we become too focused on the outputs, we will lose the very things that have enabled those outputs to exist. The inputs are the people, culture, process, data, discipline, judgment, and trust.

This September will mark fifteen years since Thrive raised its first institutional fund. Since inception, Thrive funds have compounded capital at an aggregate gross IRR of 41% and an aggregate net IRR of 33%. Each of Thrive II to Thrive X ranks in the top quartile of its applicable benchmark based on net TVPI and approximately one-third of those funds rank in the top 5% on this metric. Over the last 12 months, we have generated more than $1B of liquidity and believe there may be an opportunity for billions of dollars in additional liquidity in the coming quarters.

The reason to include these numbers is not to celebrate them, but to explain what produced them. A small number of decisions have mattered disproportionately. Each investment has taught us different lessons. Some taught us the importance of acting before something is obvious. Some taught us the value of concentration. Some taught us that the best founders can navigate periods when the rest of the world loses conviction. Some taught us that being early is less important than being right, and staying committed.

We have also made our share of mistakes. We have invested in companies that did not become what we hoped, and we have passed on companies that became extraordinary. We have underestimated products, founders, and markets. The only way to improve is to be honest about these errors of judgment and to make sure that every success and every failure becomes a data point that makes the firm better.

Our Team

We have always believed in the power of a small team. Small teams require trust, clarity, and ownership. They also require exceptionally high standards. Each person at Thrive, regardless of function, must make the firm better. We have often used the concept of an artist colony internally because we believe the best people always want to be surrounded by others who care deeply about their craft. That has been true not only on the investment team, but also across portfolio impact, finance, legal, compliance, investor relations, data, product, and engineering.

This commitment to individual and collective excellence is the source of our ability to perform for our founders and our limited partners.

Across Thrive, our teams are operating at a very high level. The portfolio impact team is helping founders diagnose the most important problems in their businesses and reason from first principles toward the right answers. Our product, engineering, and data teams now represent approximately 10% of headcount and are reimagining how Thrive itself operates. We are building tools for the investment team, investor relations, diligence, portfolio monitoring, knowledge management, and internal workflows. These internal AI tools are saving thousands of hours of work per year and critically raising the standard for what each individual and each team can accomplish.

We have always viewed Thrive as a company that happens to invest in and build other companies. In 2023, in order to think about how to best position Thrive strategically, we brought on a select group of strategic shareholders whose experience and global perspective have been invaluable to building Thrive and our portfolio companies, including Bob Iger, Henry Kravis, Mukesh Ambani, Jorge Paulo Lemann, Alexandre Van Damme, and Xavier Niel.

As Thrive has continued to evolve, some of our closest partners have expressed interest in supporting Thrive in its broadest sense. We are therefore pursuing a similarly sized minority investment from our original shareholder group, together with a small number of new institutional partners. Beyond providing additional capital to support our long-term ambitions at Thrive, these partners bring experience, perspective, and strategic counsel that will help us continue building across each of our platforms for decades to come. Our intention would be to retain the capital on our balance sheet in order to continue investing strategically in Thrive.

Across every dimension of team building and strategic positioning, we are asking the same question of ourselves that we ask of the companies with whom we partner: if we were building Thrive today, with the tools and resources now available, how would we build it? There should be no function and no process at Thrive that is protected simply because it is familiar. We need to keep shipping. We need to keep learning. We need to keep evolving and be open to new opportunities. We need to keep improving the product we deliver to founders and limited partners.

Our Commitment

Much has changed since we started Thrive. Much more will change in the years ahead. But the essence of who we are has always stayed the same.

We must continue to think independently.

We must continue to be ambitious.

We must continue to be concentrated in people and ideas, as quality is scarce.

We must continue to support founders in both their most exciting and most difficult moments.

We must continue to be disciplined when others are euphoric, convicted when others are afraid.

We must continue to be humble, as seeking external validation can distort judgment. Praise and criticism often arrive with the same lack of precision, and neither should change who we are.

We must continue to earn your trust every day.

I do not know exactly what the world will look like ten years from now. Anyone who convinces themselves that they can predict the future with certainty is not being honest. I do know that the work ahead will, at the very least, demand the same values, principles, and approach that brought us here.

We feel extraordinarily fortunate to be building during a period of such profound innovation. I believe it is in moments of extraordinary change that Thrive is able to operate at the highest possible level as our small team and generalist mandate are purpose-built to identify and execute on the most compelling opportunities.

We are grateful to the founders who allow us to be by their side. I am grateful to my colleagues at Thrive whose commitment to our work inspires me every day. And we are grateful to you for your trust, partnership, and support.

We believe the opportunity ahead is larger than anything we have seen before, and we deeply believe that Thrive today is only a fraction of what it can become.

yours,
Josh

[andrey @andreylebedev29]

@maxkarpis red bull is also a great comp for this. they created a standalone extreme sports media business to market their brand, rather than paying 3rd parties for marketing. similar play here

[andrey @andreylebedev29]

@maxkarpis very good idea, if they can keep it feeling exclusive and upmarket. vibe in existing lounges are cheapening, but those lounges are conflicted from increasing exclusivity because it will cannibalise their business-class travel business

[Prakash @8teAPi]

Incredible. Jensen is completing the circle.

- Bankers don’t like GPUs as collateral because the depreciation is unpredictable
- It’s unpredictable because a new GPU can obsolete an old one
- Jensen knows his own roadmap
- so he’s offering depreciation insurance to the banks
- the depreciation insurance (up to 25%) helps the banks get marginal deals over the line

Speculation
- Nvidia will also advise the banks on “reference designs” for datacenters that will make them fungible
- Having them be fungible means that the debt can repackaged into Asset Backed Securities, Collateralized Loan Obligations and Collateralized Debt Obligation (ABS, CLOs and CDOs from 2008 haha)
- This allows tranching to get investment grade ratings on the debt so that it can be resold to pension funds and insurance firms
- It also allows the banks to trade idiosyncratic project specific credit risk for sector wide credit risk

So Jensen is trying to get his customers the same cost of financing as real estate rather than venture equity.

This is going to move the data center game out of the VCs and into the big leagues.

[andrey @andreylebedev29]

@gm_mertd yes, totally agree, please publish! @josephinePqt1 is also doing good work on this

[andrey @andreylebedev29]

@gm_mertd the elon arc of this would be to plow all the earnings into fusion startup

[Andrew Ho @andrewho03]

I'm actually fairly bearish on frontier lab valuations. I've never seen the reasons articulated to my satisfaction, so before I go to sleep, I wanted to quickly jot down my thinking here.

The basic issue is that the labs are highly unprofitable. This may seem like a simple point, but private market valuations can be relatively irrational; however, like with $SPCX, post-IPO pricing will likely be much more punishing, especially as the standard 6-month lockup period expires and selling pressure intensifies.

Many people claim that the labs have high margins. Yet even with high margins, a valuation of $1T would be justified only if the labs were doing nothing aside from serving inference (thus reducing costs only to those relevant to inference) and posting annual revenue numbers in the $100-200 billion range assuming ~80% gross margin and a 20x earnings multiple.

This assumption is obviously not true, because the frontier labs have to continually spend money training the next generation of models. This is because of market competition from runner-up firms. For example, if OpenAI had paused model development last year, there would no longer be any point in paying GPT-5 API prices when you can just use Qwen or Kimi instead for much cheaper. Thus, the labs are forced to invest ever-increasing amounts of money in model training, in a way such that at any given point of time, the amount you're forced to invest in the next model is dramatically higher than the amount of money you're actually making, because even if your revenue goes up with higher model capabilities, so do your future training costs. This is a profoundly punishing dynamic which severely penalizes frontrunners.

(There is also a related subpoint where frontier labs claim they can distill their leading models to win out at lower intelligence levels as well. This makes no sense because the revenue numbers involved are far too low when taking into consideration the rather low margin of such inference.)

Frontier lab valuations appear largely to be based on the assumption that as you scale up, the capabilities which emerge will be sufficiently general and profound that we'll see explosive growth (https://t.co/RqmkltVpM3) from things akin to AI agents starting and autonomously managing entire companies of subagents. But it's not clear to me that this is the case; indeed, as I mentioned in my previous post (https://t.co/3URAcJ4XkJ), I believe that capabilities growth will be slower, spikier, and more data-limited than people currently assume. It may be the case that eventually we will see explosive growth of this nature with full automation of the economy, but at the very least my viewpoint implies much longer (multi-decade) timelines until we reach this point. It is not clear to me that the frontier labs will be able to operate unprofitably for so long, although I suppose maybe this foreshadows some sort of inevitable nationalization.

I also want to make a broader point about technological diffusion. The reason why technological diffusion is slow isn't just because, e.g., old people take a long time to learn how to use technology (although this is of course a contributing factor to some degree). In my view, it's because when a new, revolutionary technology comes along, the ways to incorporate that technology into subsequent developments are not always obvious, and in fact they cannot necessarily be arrived at through the application of pure reason. If they could be, then perhaps frontier models, at a certain point, would have a perfect understanding of how the LLM application layer should be developed, and they would then autonomously code, deploy, and sell such a layer.

But it seems more plausible to me that this diffusion is limited moreso by the hard problem of economic calculation--that is to say, the Hayekian notion through which the price system gradually promotes efficient allocation of resources and which cannot be simulated through central planning--and that even if we froze current capability levels at today's levels, it would take well over two decades to fully integrate in LLMs into our lives. Such a view is consequently rather bearish for the continued profitability of labs as it reduces their prospects for finding, say, something else comparable in profitability to coding agents, which seems to have been a somewhat lucky discovery by Anthropic to begin with. That is to say, even if you spam FDEs you aren't necessarily going to be able to just figure out the "correct" product shapes fast enough.

Overall, I don't think that people have clearly reasoned through their mental models for why lab equity should be worth as much as it currently is, and that if you actually bother to write down such a model, you may not arrive at the conclusion that you want to arrive at. This isn't to say that I don't expect AI to experience a huge (industry-wide) boom in the coming decades, but just that I'm not entirely sure I would buy OpenAI or Anthropic stock at latest valuations if I were given the opportunity to do so.

Of course, as an ex-lab employee, arguably this is talking against my own book; I should really be giving people more reasons to be bullish. But in the end, my influence is so small that it doesn't make a difference, so why not have some fun?

[andrey @andreylebedev29]

@MollySOShea @friedberg his argument on the inversion of income and capital gains tax thresholds is sooo well articulated and obviously the correct way for countries to operate. we need this change!

[andrey @andreylebedev29]

@gm_mertd haha true that!

[andrey @andreylebedev29]

@gm_mertd i use warp and really like it. it’s pretty non-techy imo

[andrey @andreylebedev29]

@gm_mertd yeah i think sales-led b2b is very different. and probs hard to generalise. though ive heard from the Langdock team that like close to 100% of their growth is organic inbound. crazy!

[andrey @andreylebedev29]

@0xjohnkim @paraformtalent huge congrats! 🇦🇺 represent

[andrey @andreylebedev29]

this is one of the best podcasts i’ve watched. similar to @3blue1brown. the mental model of the fastest flight path from London to SF not looking like a straight line on paper was the aha moment for me.

still wouldn’t say i fully get it, but it’s so much fun to learn about. lots of light bulb moments.

[Dwarkesh Patel @dwarkesh_sp]

Adam Brown (@A_G_I_Joe) is back!

General relativity is said to be the most beautiful idea the human mind has ever produced.

Most of us will never get to fully appreciate its elegance by taking the 20-lecture graduate course Adam taught on it at Stanford.

But in the video below, Adam distills the key idea at its heart so clearly and compellingly that even I could keep up lol.

At the core of general relativity, Einstein is trying to figure out the principle behind a particular coincidence: that the mass that resists acceleration and the mass that gravity pulls on just happen to be exactly the same. Adam then leads us through the path of insight which Einstein called his “happiest thought.”

Then Adam lectures on black holes. First, by showing how even under special relativity you could create a perpetual motion machine if black holes weren't truly black. And then, by explaining why the observations of an infalling observer and a distant bystander to the black hole would be so radically different

Adam leads Blueshift, the team at Google DeepMind cracking science and reasoning.

Which gave us the opportunity to discuss at the very end how close we are to AIs that could rediscover general relativity from scratch. Stay till the close for some philosophy of science.

0:00:00 – The coincidence that led Einstein to general relativity
0:16:42 – Gravity is a consequence of curved spacetime, not a force
0:31:46 – Why black holes prevent unlimited energy extraction
0:47:12 – Black holes are the ultimate power plants
1:13:50 – What falling into a black hole would actually feel like
1:18:51 – The three ways we know black holes are real
1:24:21 – The first time we saw gravity bend light
1:29:33 – How far can AI get without experimental evidence?

Look up Dwarkesh Podcast on YouTube/Spotify to watch. Enjoy!

[andrey @andreylebedev29]

reminds me of the deep rippling espionage situation, but like 10x worse. blatant theft of thousands of pages of trade secrets

[NIK @ns123abc]

>be Chang Liu
>senior system electrical engineer at Apple
>8 years working on iphone
>january 2026: leave Apple to join OpenAI
>apple asks for laptop back
>ignore them
>lmao it’s my laptop now
>within HOURS of leaving
>message Yu-Ting “Alyssa” Peng, friend at Apple:

Liu: “I still have another computer”

>uses it to access Apple secret info
>within weeks, use HER Apple work laptop
>february 9: try Apple’s network storage
>cloud repo of confidential engineering files
>authentication bug. still works!
>message Peng: “LOL, I found out I can access the [network storage], so funny”

Peng: “I’m ready”

>while developing hardware for OpenAI
>download DOZENS of confidential files
>including a thousand-plus-page compilation of technical files
>including MLB (main logic board) manufacturing + testing presentations

>send Peng links to Apple’s proprietary folders
>point her to specific project data
>coach her how to copy files “to avoid trouble with the security team”
>tell her which confidential Apple materials to study before her OpenAI interview
>warn her another guy “fumbled” Tang Tan’s questions about a secret Apple project
>“download some info” for her to review

>tell her: switch to LINE Messenger so nobody sees this
>she gets the OpenAI offer, leaves Apple April 16
>meanwhile every message was left on APPLE-ISSUED WORK LAPTOPS

>july 10: Apple Inc. v. Chang Liu
>named first. before OpenAI. before Tang Tan

LOL
so funny