20VC x SaaStr：英伟达万亿美元预期、Meta万人裁员与 Adobe CEO 离任

Harry Stebbings, Rory O’Driscoll, and Jason Lemkin break down the biggest stories in tech and venture right now.

Harry Stebbings of 20VC, Rory O’Driscoll of Scale Venture Partners, and Jason Lemkin of SaaStr sat down to work through the week’s biggest stories: NVIDIA’s trillion-dollar forecast, a new wave of purposeful layoffs at Atlassian and Meta, what AI fluency actually means when you’re hiring, Anduril’s $20B army contract, Travis Kalanick back at the table raising at up to $20B, and Adobe — a company with one of the most beloved product suites in history — facing a leadership vacuum and a real disruption problem.

NVIDIA: $1 Trillion … and the Market Yawned

Jensen put a trillion-dollar revenue forecast on the table at GTC. The stock moved less than 1%.

That flat reaction is the real story. As Jason put it, you could just feel the energy: “This is summer at NVIDIA. They’re on fire on everything.” New hardware. Nemo Claw, their open-source model push. A partnership with Thinking Machines. Data centers in space. The company is firing on every cylinder simultaneously.

But Harry broke down why the number itself surprised no one: NVIDIA did $215B last fiscal year, the analyst forecast for next year was already in the mid-300s, and the “trillion dollar” headline was essentially a salesman’s round-up of two years of consensus estimates already priced in. The stock processed it in ten minutes.

What the announcement actually confirmed is the bigger bet: capex investment at these levels continues for four to five more years. When NVIDIA is doing $600B in revenue, the global capex spend behind it is probably north of $1.2 trillion. Jason’s view on the longer arc: cumulative revenue of $10 trillion over the next five to seven years is plausible, driven by inference running at scale across a growing share of the global workforce.

Rory flagged the one real tension in the math: token consumption may grow 3,000x over the next five years, but if the price per token falls 6x simultaneously, revenue growth is not linear. The bull case on NVIDIA requires demand to outrun price compression at a massive scale. So far, every data point says it is. But it’s a bet worth naming as a bet.

The counter-risk is real but so far theoretical. Inference prices keep falling. Customers are building on Google TPUs and Amazon chips. And railroad-boom-level capex assumptions require nothing to break for half a decade. Harry put the probability of something going wrong at around 30%. Not enough to bet against NVIDIA — but worth noting.

The Layoffs at Atlassian and Meta Are Not What They Look Like

Atlassian cut 1,600 people. Meta is reported to be heading toward 16,000 out of 79,000 — roughly 20% of its workforce. The coverage frames this as AI forcing companies to downsize. That framing misses what’s actually happening.

As Jason said plainly: “Neither of these companies has to lay off anybody.” Atlassian has substantial free cash flow. Meta’s operating margins are still in the 40s. These are not distress moves.

Rory laid out four categories of what’s actually driving this wave:

• First, companies that overhired in 2021 and are now correcting — using AI as a convenient explanation, but the efficiency metrics never justified the headcount.
• Second, businesses that were growing at 20% and are now growing at 2%, and need to give Wall Street what it wants. As Rory said: “Wall Street is simple. If you give them growth, they leave you alone. If you don’t give them growth, you better give them profitability. And if you don’t give them either, they’re going to bust your chops.”
• Third, companies where AI tooling has genuinely replaced functions that previously required humans — coding being the clearest example right now.
• Fourth, and this is the Meta case specifically: you spent tens of billions on compute. The depreciation is coming. You don’t have the operating cash flow to have both Nvidia and people. As Harry put it, “compute eats jobs.” That is literally what is happening at Facebook.
• The fifth category — which Jason added — is arguably the most important and most underappreciated: companies cutting not to shrink, but to restock. They don’t need 20 engineers who know C++. They need eight who are genuinely elite at building with AI. They’ll pay twice the salary for half the headcount. This is a talent shuffle happening in real time, and it probably should be happening at every company regardless of whether any of the first four factors apply.

The One Question That Tells You If Someone Is Actually AI Fluent

Harry asked Jason for a concrete hiring test. The answer was blunt.

“What commercial AI tool have you brought into your organization this month? That’s the test.”

Not which tools they’ve read about on Twitter. Not which demos they’ve watched. What did they actually buy, configure, deploy, and put in front of their team — in the last 30 days. Anyone on the bleeding edge has done this repeatedly. There are enough great products now that there’s no excuse for any functional leader — sales, marketing, engineering, product — to not have evaluated and partially deployed at least one agentic tool recently.

Jason was direct about where the bar actually sits in practice: “Of all even of the best startups I’ve invested in, maybe 30% of the management team meets this standard at best.” In general interviews, it’s single-digit percentages.

The role that matters right now isn’t a prompt engineer — that job existed for about a year and is already gone. It isn’t “go-to-market engineer” either, a title Jason thinks is also fading. The job is what Jason coined on the call: agentic deployment expert. Someone who finds the best tools, deploys them, trains them properly, and measures the output. From C-suite to junior.

Rory and Harry pushed back gently — anyone can learn this, provided they’re willing to put in the time. It isn’t a generational filter or a technical one. Harry made the point explicitly that it applies to investors too: if you’re not using these tools yourself, you probably shouldn’t be writing checks into them. Harry called Rory out directly as the most AI-equipped investor he knows — sitting genuinely at the bleeding edge in terms of using the tooling, understanding where the bottlenecks are, understanding constraints. That’s what good looks like on the fund side.

And critically: you do not need to be technical to do this. “You do not need to be even 1% technical,” noted Jason. If you’ve successfully deployed a piece of enterprise software — Salesforce, HubSpot, Outreach, anything — in the last few years, you have the skills to deploy AI agents. The only genuinely non-intuitive part is training them. Everything else is just software. The people who figure this out now become heroes in their companies. The people who don’t, as Jason said, are “going to get laid off because they’re not going to matter.”

Anduril Just Became the $20B+ Default OS for the U.S. Army

Anduril landed a $20B, 10-year army contract — five-year base with a five-year option — consolidating 120 separate procurement actions into a single enterprise agreement.

The structure is what matters. This isn’t a new program; it’s the army saying “you’re our prime supplier, let’s stop doing 120 separate contracts and just systematize this.” The product at the center appears to be Lattice, Anduril’s software connectivity layer that ties disparate hardware systems together in real time.

In modern conflict, you have seconds to respond to an incoming drone. You don’t have time for a slow connectivity protocol and you definitely don’t have time for a human in the loop. The military needs all its offensive and defensive systems talking to each other autonomously. Anduril’s product is becoming the default layer for that. The Pentagon has essentially said “you’re the dominant provider of this layer — let’s systematize the contract.” It picks them as the clear new prime.

For B2B founders: this is the integration layer playbook. Become the connectivity fabric across a complex multi-vendor environment. Let everything else plug into you.

Seed Investing Is Broken (Again). Growth Funds Are Winning.

Jason made a confession that started a spirited back-and-forth: he’s essentially given up on his prior thesis that a small TAM with a great founder can stair-step into a big one. “I can’t even bring myself to take a meeting with a startup where I don’t believe the TAM will be utterly massive.”

Harry pushed back — and then offered a sharper framing of the real problem. It’s not just about TAM size. It’s that YC has productized the $60M post-seed valuation. When you’re paying power-law prices for mid-tier market opportunities, the math doesn’t work. You need something like a 250x on a $60M post to return a meaningful fund in today’s environment. Harry’s question: how many public tech companies have market caps north of $15B? Fewer than 50. The math is grim.

The counter-argument is TAM velocity — the speed at which a market expands — which is distinct from TAM size at entry. The Anduril example is instructive: if you’d looked at “watchtower software for the US border” as a TAM, you’d have passed. If you’d looked at “the connected command layer for all of US defense procurement,” you’d have funded it. Same company, same product, very different TAM frame.

Harry’s conclusion: “The growth fund strategy is currently winning.” Wait for proof. Invest into demonstrated product-market fit. In an environment where AI is creating genuine uncertainty about which markets expand and which get destroyed, actual revenue data beats a compelling deck.

That said, as Harry noted, excess capital tends to compress returns at any stage once a thesis becomes consensus. “There’s no investment opportunity so good that excess capital won’t destroy it” — a Barton Biggs line Harry invoked more than once.

Travis Kalanick Is Back. The Question Is Whether the Opportunity Matches the Founder.

Travis Kalanick, eight years after being forced out of Uber by the board, re-emerged with Atoms — an industrial robotics company building on what he learned running Cloud Kitchens. A thousand employees. The company also rebranded City Storage Systems. His thesis: robots on wheels, not humanoids, are the near-term opportunity.

The robotics call is sharp. Humanoid legs consume battery life, introduce instability, and are unnecessary for the vast majority of factory and logistics applications. Wheels are more efficient, more reliable, and more than sufficient. As Jason noted, even Sunday, the home robotics company that recently raised, landed on wheels too. Travis’s call implicitly bets against the current humanoid hype cycle — not saying it never happens, but that the path probably runs through purpose-built, non-humanoid machines first.

The harder question is the Uber counterfactual. Jason was direct: “If he were running Uber today, it would be a trillion dollar company. Without question.”

His argument is that Travis’s hyper-aggressiveness — the trait that ultimately got him removed — destroyed Lyft, and in today’s era of speed-as-strategy, it would have been the decisive asset rather than a liability. He would have dominated food delivery with capital and market share instead of dribbling into it through acquisition. He would have been five years ahead of where Uber currently sits on autonomy.

Harry offered a more nuanced read: there was likely a real period — roughly 2017 to 2022 — where a financial-operator CEO was actually the right person for Uber’s stage. The company needed to get public, get cash-flow positive, and give Wall Street the profitability story. Travis was genuinely unwilling to do that. Replacing him may have been the right call for that chapter, even if it cost the company long-term product velocity. “Maybe they should have done the Steve Jobs thing — swapped him out, got it public, got cash flow positive, and brought him back in 2022.”

On whether to invest in Atoms at a reported $20B valuation: Harry’s view was clear and fund-size-dependent. “If I’m General Catalyst raising $10 billion, do I want to put a couple hundred million into Travis, one of the greatest founders of our time? Yes.” Rory was equally direct from his position managing a large growth fund — if you have $4.5 billion a year to deploy, putting a few hundred million behind one of the best founders alive is a defensible swing. You’re not betting the fund. Jason’s view was the inverse: at his fund size, he needs 100% conviction — on the founder’s current hunger, on the specific opportunity, on the market — and the interview gave him some hesitation. “At some point you do lose it. You lose the ability to create.” His read: not there yet necessarily, but the interview had some notes that felt like 2017 rather than 2026. As Harry summarized: “Fund size is strategy.” The bet looks different depending on what you’re managing.

Adobe’s CEO Stepping Down. It May Have a Real Problem

Adobe beat earnings. The stock dropped. And the CEO, Shantanu Narayen, announced his departure simultaneously — without naming a successor.

That sequencing is the tell. Standard transitions have the departing CEO identify their replacement and announce an orderly handoff. When you announce you’re leaving before you have the answer — especially after 18 years — something went differently than planned. The most likely explanation, as Harry and Jason worked through it: either the board decided it needed to send a signal to preempt activist pressure (Elliot-style), or Narayen walked in and said his heart wasn’t in it anymore.

Neither is confidence-inspiring.

The structural problem at Adobe is more important than the leadership question. Adobe’s aggregate stock returns over the last decade are roughly in line with the S&P 500. For a company with this brand, this installed base, and these products — Photoshop, Premiere, Illustrator, the full Creative Cloud — that is underperformance. The moat has held. The growth hasn’t materialized.

The AI risk is specific and acute. As Harry framed it: “What kills you as a software business is if the work you’re automating gets done in a totally different way.” Adobe’s core is the professional creative workflow. That is precisely what AI is rewriting. When a marketing team can generate a production-quality image in seconds, the calculus on an Adobe subscription changes — not immediately, not completely, but directionally and irreversibly.

Jason compared this to Intuit and didn’t like the comparison for Adobe’s benefit. Intuit’s defensibility comes from regulatory complexity and data history: accounting rules, tax codes, years of a business’s financial data locked in the system. Adobe’s defensibility comes from professional skill. And AI is eroding the value of that skill in the consumer and mid-market faster than most people expected. As Jason put it: “I see no evidence Adobe will grow. I see no products that show they’ll grow. Nothing.” Five years from now, we’ll still be filing taxes. We may not be doing pixel-by-pixel work in Photoshop.

The Bottom Line

• NVIDIA: The infrastructure bet has been won. The trillion-dollar forecast was priced in. The question is what gets built on top of $10T in cumulative compute spend.
• Layoffs: These are purposeful decisions, not distress moves. The real story is the talent shuffle — companies aren’t shrinking, they’re swapping for AI-fluent people at higher salaries.
• Hiring: One question tells you what you need to know: what AI tool did you deploy this month? If they can’t answer it specifically, don’t hire them.
• Anduril: Owning the integration layer in a complex multi-vendor environment is the most defensible position in any market.
• Travis / Atoms: Elite founder. Real robotics thesis. Hard math at $20B unless you’re managing a fund that has to move hundreds of millions a month.
• Adobe: A 30-year head start that may be running out of time to use it.

Quotable Moments

Jason Lemkin:

“The job today is not to screw around on ChatGPT and be a prompt engineer. The job today is to bring the best agentic products into your organization. You have to be an agentic deployment expert. From C-level to junior. Don’t hire anybody else. You’re going to regret it.”

“You do not need to be technical to win with AI agents in Q2 of 26. You do not need to be even 1% technical. This is your time. Or you’re going to get laid off because you’re not going to matter.”

“If Travis were running Uber today, it would be a trillion dollar company. Without question. That guy would be running a trillion dollar company today if Gurley and team — and importantly, his own actions — didn’t force him out.”

Harry Stebbings:

“Wall Street is simple. If you give them growth, they leave you alone. If you don’t give them growth, you better give them profitability. And if you don’t give them either, they’re going to bust your chops.”

“Today, compute eats jobs. You literally can’t afford to have Nvidia and people.”

“You should be very, very, very wary of ever swapping out a founder. It’s like open heart surgery and 50% of people die. It is easier to just lose money.”

Rory O’Driscoll:

“You could have 3,000 times more tokens, but if the price per token goes down by 6x, the revenue would decline. Just pointing.”

“We want big picture trends and near traction. We don’t want to compromise between the two. Something has to be working right now, and you have to be able to articulate the expansion story.”

“Whenever you find your logic being reduced to: I’ve got to move this much money this month, the entrepreneur is amazing, I’m not sure about the opportunity, I don’t like the price, but have a go — that’s when you should think hard about your fund size.”