Gavin Baker Just Bet $370 Million on a Two-Year-Old Chip Company. I Ran My Own Numbers.

Astera Labs went public 25 months ago. It trades at 116 times next year's earnings. One customer, widely believed to be Nvidia, generates more than 70% of its revenue.

None of that stopped Gavin Baker from nearly doubling his position in a single quarter.

I run a Compounding Portfolio built on eight or nine businesses I plan to own for decades. Arista Networks and Amazon are my only two winners this year. Both ride the same wave: AI clusters need networking infrastructure to actually work, or the GPUs inside them are just very expensive space heaters. So when a superinvestor like Baker builds a $370 million position in a company most people have never heard of, I go find out why. Here is what I found.

What the company actually does. Picture a rack of Nvidia GPUs wired together to act as one machine. At the signal speeds those GPUs now run, electrical data degrades within inches of copper. Astera Labs builds the chips that catch that signal, clean it, and pass it on. Retimers. Smart cables. PCIe switches. A software layer called COSMOS that lets hyperscalers watch link health across an entire data center floor. Nvidia's own reference designs for its HGX and MGX server platforms have shipped with Astera silicon since 2020.

The numbers, cold. Revenue hit $308 million last quarter, up 93% year over year. Non-GAAP operating margin: 36%. Free cash flow margin: 34%. Zero debt, $1.2 billion in cash. This is not a story stock burning capital to grow. It is a real, profitable, fast-growing business.

Here is where I slow down. Astera trades near 44 times next year's sales and 116 times next year's earnings. Wall Street's own price targets, even after raising them twice this year, average roughly 30% below where the stock trades today. The market is already pricing in years of flawless execution. That is not a moat conversation. That is a bet that the multiple stays rich.

The moat, stress-tested. A retimer is not CUDA. It is not AWS. It is a socketed chip that Nvidia, or any hyperscaler, can re-qualify a different vendor into on the next board redesign.

Broadcom and Marvell already control more than 80% of the broader connectivity market at gross margins above 65%. Credo, a smaller direct competitor, grew revenue 157% last quarter. Faster than Astera. Off a smaller base. And reports around Nvidia's next Blackwell Ultra platform suggest at least some configurations may not need a standalone retimer at all, because part of that function is moving onto the GPU's own package.

That is the real risk. Not that AI infrastructure spending slows down. That Astera's specific slice of it gets designed out of the next generation.

Customer concentration. One customer, widely believed to be Nvidia based on the product mix (Astera does not name it), drove more than 70% of 2025 revenue. The top three drove 86%, up from 80% the year before. Concentration is getting worse, not better, as the company scales. One re-design decision by one customer moves this stock more than any macro headline will.

Why Baker bought it anyway. Gavin Baker runs Atreides Management, a fund built around one idea: in the AI buildout, value accrues to whoever owns the scarce physical inputs, not the software sitting on top of them. His public framing of Astera argues the market miscategorizes it as a "copper loser" in the copper-versus-optics debate, when its actual product line spans both. He added over a million shares last quarter alone, pushing his position to $369 million, roughly 7% of his disclosed portfolio and his largest non-index holding. That is real conviction, sized like a hedge fund bet. Not a buy-and-ocation.

Why it is not Arista or Amazon. Arista has a decade of profitable operating history and a software lock-in, EOS, that keeps network engineers loyal for years. Amazon's AWS moat is built on switching costs measured in the cost of re-architecting an entire company's infrastructure. Both trade at a fraction of Astera's multiple. Astera Labs is two years old as a public company. It has never been tested through a single capex downturn. Riding the same theme as Arista and Amazon does not make it the same kind of business. It is an early, unproven, single-customer-dependent bet wearing the same jacket.

Five scenarios, not one number. A casino does not predict the next hand. It prices every hand. Here is how I price this one over three years, from today's roughly $400 share price.

| Scenario | Revenue CAGR | Net margin | Price target | Return | |---|---|---|---|---| | Super Bull | ~65%/yr | 40% | ~$540 | +35% | | Bull | ~40%/yr | 33% | ~$210 | -47% | | Base Case | ~28%/yr | 28% | ~$115 | -72% | | Bear | ~12%/yr | 22% | ~$43 | -89% | | Super Bear | revenue shrinks | 15% | ~$20-25 | -94% |

Read that table twice. Even the Bull case, real 40% annual growth sustained for three years, loses you money at today's price, because the multiple has to come back to earth. Only the scenario where Astera keeps growing at close to its current insane pace, and the market keeps paying a Nvidia-in-2024 multiple for it, makes you money from here. That is a narrow lane.

The scorecard. I score every name in this portfolio the same way, so you can compare it to anything else I cover.

| Dimension | Score /10 | |---|---| | Essential or nice-to-have | 8 | | Current moat | 4 | | Moat expanding | 4 | | Balance sheet strength | 9 | | EPS accelerating | 9 | | Net margins increasing | 7 | | ROIC profile | 7 | | Reinvestment rate | 7 | | Capital return | 2 | | Valuation (reverse DCF) | 2 |

Overall: 59/100. A genuinely excellent business. Not yet a genuine compounder. Baker is paid to take concentrated, early bets on physical scarcity. I am not. I will keep watching Astera Labs. I am not buying it at 44 times sales with 86% of revenue sitting behind three doors.

That is the discipline. Five scenarios and an honest scorecard, before a dollar moves. It is the same process I teach inside Sprint Club.

Start your free 7-day trial and build your own scorecard on the next stock that shows up on your feed: https://www.strategysprints.com

Happy hunting. Simon & The Sprinters 🐬⚡️🐆