With Harry Stebbings, Jason Lemkin, and Rory O’Driscoll
Apple sued OpenAI for trade secret theft this week, and the details are ugly: a six-year Apple employee allegedly walked physical prototypes out the door for show-and-tell, encouraged by a 24-year Apple veteran now running OpenAI’s hardware group. The individuals involved are, in Jason’s words, toast. But the more interesting read is what the lawsuit signals about OpenAI’s hardware ambitions, which look increasingly like a distraction from the one thing actually printing money right now: coding.
That tension ran through the whole episode. Meta re-entered the frontier race by launching Spark 1.1 and, for the first time ever, charging developers for its own models. Zuck broke a three-year silence on X to do it. The battle is aggressively priced and aimed squarely at the cheap-token tier, because every company with a half-awake CIO is about to hand its developers a budget model to stop the bleeding.
And the bleeding is real. ClickHouse’s AI spend is up 60x since February. Jason burned an entire month of Claude Design credits redesigning a single page. The best developers are now coding 24 hours a day across 10 to 20 agents, and none of them can consume enough tokens. Which raises the question Rory kept circling: there are only 1.8 million developers in the US and roughly $250 billion in total developer wages. If most of Anthropic and OpenAI’s enterprise revenue is coding, they may already be at 20% of the entire software labor market. Nobody on the pod had a conclusion, but the possibility is real: the fastest-growing companies in history could hit the ceiling of the till faster than any company ever has.
Top Takeaways #
1. Apple vs. OpenAI Won’t Slow the Roadmap
The facts are damning for the individuals. The junior employee who took the material did it, there’s fact-based evidence in the 41-page complaint, and he’s already lost his job. The first thing he’ll learn is that there’s no gratitude in litigation: the moment it’s convenient, OpenAI burns him. The senior exec who allegedly encouraged the hires faces a harder discovery process, and Apple has every incentive to twist the knife up the chain. Apple is furious that 400 of its people have gone to OpenAI, and now they have leverage. But none of it changes the ship dates. It’s a message.
2. California’s Employment Law Made the Theft Pointless, and Anthropic Is the Proof
This is what makes the lawsuit look self-inflicted. California doesn’t enforce non-competes, and the doctrine of inevitable disclosure means the knowledge in your head is yours to carry to your next employer. Anthropic is the single biggest beneficiary of that legal reality. When those founding employees walked out the door, they didn’t need to take a single line of code or a piece of paper. Everything in their brains was already theirs. If you can hire the domain expert legally, why put yourself at criminal risk to steal what you were already allowed to use?
3. OpenAI’s Hardware Bet Looks Like It’s on the Cutting Board
The signals point one way. Hardware made sense when OpenAI had an unassailable consumer lead and everything was working. Now Sora’s been trimmed, the TBPN acquisition looks questionable, and the $6 billion for Jony Ive’s team just produced a lawsuit instead of a product. It’s starting to resemble the Apple Car: a cash-hemorrhaging distraction that keeps getting closer to the axe. This lawsuit could be the thing that pushes leadership to it for a year. As a pure consumer company, hardware and media bets made sense as concentric circles. As a company that has discovered the real value is enterprise coding, it’s a death march.
4. Meta’s Spark Launch Is a Direct Assault on the Cheap-Token Tier
Meta shipped Spark 1.1, moved off open weights, and started charging on an API, adopting the exact business model as every other frontier lab. Alexander Wang called the pricing very aggressive versus OpenAI and Anthropic. The product is good enough as a consumer experience, but the real test is coding, and their benchmarks have historically overpromised. The strategic point holds: Meta is entering with a low-priced product aimed at the second bucket of demand. Whether it’s ROI-positive to be the fourth player selling at that price over five to ten years is an open question.
5. Every Company Is About to Get a Two-Tier Model Strategy
Token-maxing has gone from niche to inevitable. Everyone is going to hit their limits, and every organization is moving from “burn all the tokens you want” to budgets, whether that’s enforced by automation, a dropdown, or a manual cap. That creates a permanent slot for a cheaper model, either inside your product for simple workflows or for lower-stakes coding. Anthropic’s Haiku got specific praise here: for cheap, simple work it’s excellent and costs a tenth of a cent, and Anthropic could make it better overnight just by adjusting how much Opus reasoning flows into it. The frontier fight gets the tweets, but the volume battle is happening one tier down.
6. The Databricks Paper Kills Cost Per Token: The Metric Is Cost Per Completed Task
Cost per token is not a useful metric. The number that matters is cost per completed task, because a model with cheap base tokens can have expensive reasoning tokens and unpredictable consumption, blowing past a nominally pricier competitor. From there you get a Pareto curve: different models win for different task types, and a cheap model that requires infinite reasoning can lose to paying up for the expensive one. The third finding: the harness and infrastructure around the model massively impact efficiency. Yes, Databricks is a company that sells model management telling you model management matters. They also produced the data to back it up.
7. AI Token Budgets Are Exploding, and the Richest Workflows Haven’t Even Started
ClickHouse’s AI spend is up 60x since February. That’s not an outlier once you do anything complex. Instead of redesigning one page, have Claude Design redesign your entire 100-page site every night so you wake up and pick the two or three better versions. That’s orders of magnitude more tokens, and it’s a better outcome. Every top-tier developer and designer could consume 100x what they use now without blinking. Jason’s real-world version: he burned a full month of Claude Design credits on a single page redesign, and the workflow he actually wants would cost multiples of that.
8. Losing the Bottom of the Funnel Is the Real AI Risk for Incumbents
Claude Design won’t kill Figma, and it won’t show up in a quarter’s numbers, because it’s taking the ankle-biter, single-seat deals at the bottom. But those ankle-biters move up the leg over time. If you lose the bottom of the funnel, the next generation may never graduate to you, because they started agentically and never needed you. This is where Salesforce is exposed too. The only metric that matters for a public B2B company here: are they still growing net new logos 15% or more a year? If yes, they’ll figure it out. The moment that falls, price increases only carry you so far before the death spiral.
9. AI Coding Revenue May Already Be a Fifth of the Entire US Developer Wage Bill
The BLS math is sobering. There are only 1.8 million developers in the US, of which roughly 200,000 work at software companies and another 600,000 at tech companies. Total developer wage spend is around $250 billion. If a large share of Anthropic and OpenAI’s enterprise API revenue is coding, they may already be close to 20% of that entire market, without the layoffs you’d expect to see. Rory was careful not to call it: there’s no hard evidence they’re hitting a limit, and the momentum says the opposite. But the arithmetic is worth sitting with. Things growing 10x don’t stop cold, yet even 6x next year off a $50-60 billion base is $300 billion, more than the total US developer wage bill. Either coding revenue decelerates hard, or a wall of CFOs is about to ask why tech spend doubled while headcount didn’t move.
10. The Other TAM Vector: A 10% Software Tax on the Whole Industry
There’s a second math worth running. Software spend is roughly $1.4 trillion, and it’s all going agentic fast. If companies tolerate a 10% gross-margin spend on AI, the same way they accepted a 7% Amazon tax a decade ago, that’s another $100-140 billion accessible to Anthropic and friends. Outside of coding, agentic token costs are manageable, closer to 10% than 50%. Salesforce could comfortably pay 10% of revenue in tokens. It won’t pay 40%, because it only runs 22% operating margins. So there are three buckets: coding, the 10% agentic software tax, and eventually co-work replacing the knowledge worker at Office-like price points.
11. SK Hynix’s $26.5B NASDAQ Listing Exposes a Brutally Cyclical Business
SK Hynix priced the largest-ever NASDAQ listing by a foreign company, popped 13%, then gave it back. The three memory companies (SK Hynix, Samsung, Micron) have been massive AI capex beneficiaries, with SK Hynix up 6x and operating margins swinging from negative in 2023 to 70% now. The bull case: they trade at 5 to 8 P/Es, dirt cheap. The bear case: memory is historically capital-cyclical, and 70% net margins don’t survive new capacity coming online. IBM’s stock crashed 20% this week and blamed memory directly, saying CIOs are scrambling to buy it before prices climb, leaving nothing in the budget for mainframes. Probably an excuse. Probably also a germ of truth.
12. Late-Stage Venture Is a New Asset Class, and Calacanis Pivoting to Growth Proves It
When Jason Calacanis, who spent decades building an early-stage syndicate machine, announces he’s tilting entirely to later-stage growth, it’s a sign of the times. The structural story: the early-stage business hasn’t grown much in two decades, but a 5x larger late-stage business has emerged as companies stay private and scale to billions in five years. This is a new category, private $100 million+ checks into companies already worth a billion, replacing what the NASDAQ small-cap market used to do. Layer on the most liquid secondary market anyone’s ever seen, and now is structurally better than prior cycles. But comparative advantage still matters: a great first-check investor doesn’t automatically monetize that skill at $20 billion pre.
13. Constellation Buying TouchBistro at 1x Shows What Stalled Pre-AI B2B Now Sells For
Constellation bought TouchBistro, a $70 million ARR Toast competitor, for $70 million. One times revenue. The mechanics: venture debt from Francisco Partners converted to senior equity when the company missed, wiping out common. Jason’s position on debt is unambiguous: don’t take on a wall of debt on top of a slow-growth business, and never do debt instead of an equity round. But the valuation is the real signal. No egos, no 2021 markups to defend, no founder drama. Just a stalled unicorn at 6-8% growth with deteriorating share, priced at exactly 1x. In the agentic world, these can’t be turned around, and the terminal decay is accelerating faster than expected. Marketo is the case study: no net new logos, everyone hates it, and Salesforce’s LLM-powered migration cut the switching cost from a year to a single day.
Quotable Moments #
Harry Stebbings
“Isomorphic was oversubscribed 8 to 10x. The majority of people who wanted huge amounts got nothing at all. That creates a blast radius, where suddenly there’s a lot of interest in Chai, and then a lot of interest in Latent Labs in London. It’s a blast radius from one very hot deal.”
“It really depends who says it. When I say that, I get in so much trouble. And then Paul Graham and YC say it and it’s ‘Yeah, marvelous.’ No one’s scared of me, that’s the difference.”
“The rise of tranched rounds isn’t new either. It’s just exacerbated by AI. These things aren’t brand new, they’ve just become normal for potential outliers, and there are just more potential outliers now.”
Jason Lemkin
“It’s 10 at night, and instead of doom scrolling, I’m doom coding. I go all night, and in the morning I want to check in on that workflow. I don’t know any AI developer who couldn’t consume even more frontier tokens. We just want more.”
“Marketo is going to zero. There are no net new logos, everyone hates it, and they just raised prices on us 20% again while deprecating the API. Salesforce did an LLM lift and it took one day to leave. It used to take a year.”
“If you don’t want to be worth 1x, do something before it’s too late, man. A slow-growth company at 6-8% with a big wall of debt on top of it? That’s your future, and it’s a brutal lesson.”
Rory O’Driscoll
“In the early stage, when you’re on the board, you can’t invest in two competitors. In the late stage, structurally, you have to. Fidelity Growth would have invested in both OpenAI and Anthropic. They’re not on the board of either. They want to make the secular bet.”
“It may well be that the prize for becoming the fastest-growing company in human history is that you hit TAM faster than any company in recorded history. You may have hit the limits of how much money there is in the till.”
“Treason doth never prosper. What’s the reason? For if it prosper, none dare call it treason. When you succeed as Uber, you get what you did legalized. When you succeed as Airbnb, you get what you did legalized. Cookie stuffing is a little less likely to clear that bar.”
This post is part of the ongoing 20VC x SaaStr collaboration with Harry Stebbings and Rory O’Driscoll. The My Takes companion posts Sunday.