Power 20 Q1: OpenAI -22%, Anthropic +59%

Augment just published its latest Power 20 rankings, updated for Q1'26. The headline number is staggering on its own — the 20 most active pre-IPO names on the platform now carry $4.25 trillion in aggregate estimated market cap, up roughly $940B in a single quarter, larger than Amazon's public market cap and 2.5x Meta's.The story underneath the headline is weirder. Q1'26 was the quarter OpenAI posted its first 22% drawdown in Augment history — in the same ninety days it closed the largest primary funding round in Silicon Valley history. Anthropic, meanwhile, gained 59% on the secondary market and is now fielding unsolicited offers at $800 billion, more than double its February primary mark. Two AI leaders. Opposite directions. Same quarter.

‍

The Power 20's First Real Divergence

For most of 2024 and 2025, private AI traded as a block. Every round at every leader repriced every other leader. OpenAI's SoftBank round marked up Anthropic. Anthropic's Amazon expansion marked up OpenAI. The secondary market, such as it was, mostly agreed with the primary market because there wasn't enough trading volume to disagree. "AI" was the factor. That was the trade.

Q1'26 broke the consensus. Between January and March, OpenAI's Augment Price fell 22.0%, from $881.57 to $687.68 — the largest single-quarter drawdown in Power 20 history. In the same quarter, the company closed a $122 billion primary round at an $852 billion valuation, the biggest private financing ever recorded in tech. The secondary market is now trading OpenAI as if it were worth roughly $765 billion — an appreciable discount to its own primary mark, with about $600 million in OpenAI shares reportedly sitting on offer without takers.

Anthropic went the other way. Up 59.4% on the secondary. Activity up nearly 3x quarter-over-quarter — the first quarter in Power 20 history that any name displaced SpaceX as the most-active ticker. By mid-April, Bloomberg reportedAnthropic fielding unsolicited investor offers at $800 billion-plus, more than double the $380 billion mark set in its February Series G.

Two AI leaders. Opposite directions. Same quarter. That is not noise. That is dispersion.

The evidence widens from there.

The top five names in the Power 20 averaged 20.2% price gains in Q1'26, down from 46.2% in Q4'25. The bottom 15 averaged 29.9% — the first quarter since Q2'25 when bottom-half price performance beat the top. Four names printed outright negative quarters: OpenAI at -22%, Canva at -16%, Rippling at -11%, Perplexity at -5%. A quarter earlier, only one name was negative.

And yet activity concentration hit an all-time high. The top six names — Anthropic, SpaceX, Anduril, Stripe, Polymarket, Databricks — accounted for 71.9% of all Power 20 activity, up from 58.0% for the same cohort in Q4'25. Across the entire Augment platform, those six commanded 47% of bid-ask volume, up from 40% the prior quarter and roughly 25% just two quarters earlier. Total platform activity rose 81% QoQ. The top-6 share grew faster than the platform itself.

Here is the part consensus misses

Concentration and dispersion are not opposites in the Power 20. They are both signs of the same underlying event: a market that is starting to price names against each other rather than as a block.

When private markets get enough liquidity to do real price discovery, names stop trading purely off primary marks and start trading off supply and demand. When a primary round frontran the secondary — as OpenAI's rolling tender offers and SoftBank-led rounds did through late 2025, with investors pulling forward expected mark-ups — the secondary has to give back. When the primary mark lags the fundamentals — as Anthropic's $380B February mark arguably did against a reported $14B revenue run-rate growing 10x annually — the secondary runs ahead.

That is position-level relative value. The defining behavior of a public market, showing up in private. It isn't new. It just hadn't surfaced in private markets because there wasn't enough trading volume to resolve it. Q1'26 had the volume. Activity across Augment rose 81% QoQ; within the Power 20, aggregate activity rose 76%. At that resolution, the market has enough signal to say one AI leader is mispriced against another — and it did.

The dispersion is the signal. The concentration is how that signal gets expressed — capital pools into the names where institutional buyers and sellers disagree enough to trade.

What this means for secondary investors.

Three implications follow from Q1, each structural rather than directional.

First, for the first time in Augment history, Anthropic is trading at a premium to OpenAI on a secondary basis. Buyers are stating the logic openly: "People are betting that Anthropic's valuation will catch up with OpenAI's." You can no longer buy "AI exposure" and expect the names to move together. The decision now is about which AI names, at what marks, against what fundamentals.

Second, primary rounds are losing some of their price-anchoring power. Historically, a new primary round dragged the secondary with it. Q1 broke that. OpenAI's $852B primary mark coexisted with a 22% secondary decline. Databricks' $134B February mark coexisted with only a 10% secondary gain. Primary rounds increasingly set ceilings; secondaries now set floors. Which one is "the market price" is becoming a question rather than an answer.

Third, the bottom-15 breadth is a separate signal from mega-cap rationalization. The Q1 standouts outside the top tier — Cerebras (+173%), Shield AI (+61%), Saronic (+57%), Neuralink (+55%), Kalshi (+47%) — each had a concrete catalyst. Cerebras filed for a Nasdaq IPO on April 17 at a $22-25B target. Anduril's $20 billion Army contract vehicle pulled the entire defense-tech cohort. Saronic closed a $1.75B round at $9.25B — more than doubling its 2025 mark. Breadth up, coupling down, signal-to-noise up.

The counterargument.

One quarter of data does not make a regime change. The last two times the bottom 15 outperformed the top 5 (Q2'24 and Q1'25), the top reasserted itself the next quarter. A single catalyst at a single name (Cerebras's +173% print on ~3x activity) carries disproportionate weight in the "breadth" story. Not every price move is a clean signal — Cerebras trades on thinner liquidity than OpenAI, and Augment's own methodology acknowledges that wider bid-ask spreads follow lower activity. The rotation looks cleaner on the scoreboard than in the order book.

There is also a structural distortion. xAI exited the rankings after its February merger into SpaceX. xAI was Q4'25's #7 name and one of the strongest performers at 73% that quarter. Removing it mechanically lowered the top-5 average and, to a smaller degree, raised the bottom-15 average. Without that base-effect, the breadth signal thins.

None of that negates the core observation. Even adjusting for xAI, even discounting thin-market names, the OpenAI/Anthropic split is real, reported, and widening. The harder question is whether the market stays rational about the split — or whether Anthropic, now priced at a premium, simply becomes the next name for primary rounds to frontrun and secondaries to reprice.

📈Data Point of the Day

$4.25 trillion

The Power 20's aggregate estimated market cap at the end of Q1'26, up roughly $940 billion (+28%) from $3.31T in Q4'25. For scale: that exceeds Amazon's current public market cap and runs about 2.5x Meta's. Twenty private companies, one Amazon-plus.

🎓 Manual

Primary Mark

The valuation struck in a company's most recent priced funding round — set by a lead investor, the company, and the round's lawyers. Primary marks are the reference point most investors cite for a private company's "value," but they aren't live prices. They reset only when a new round closes, which can be 12 to 18 months apart. Secondary marks, by contrast, reprice continuously as buyers and sellers transact. When the two diverge — as OpenAI's did in Q1 — the gap is the clearest read on how much the market has learned since the last round.

‍