Weekly Pre-IPO Stock Secondary Market Update | November 10, 2026

VAST Data reportedly has 14 customers committing more than $10 billion to VAST-powered infrastructure in 2026, including four customers committing more than $1 billion each, and we think the company is getting too big to ignore. VAST works in the software layer of the AI stack, helping customers get more out of memory and storage rather than making the hardware itself. Memory hardware has been one of the strongest public trades, with Micron and SanDisk rallying sharply over the past year and a half, and VAST gives investors a way to approach the same bottleneck from the software side. The technology is highly specialized, which makes the business hard to evaluate at a deeper level, as with most deep AI infrastructure companies. Even so, customers committing more than $1 billion each shows how much sophistication it takes to deliver AI at scale, which we think most investors do not fully appreciate. VAST is not yet under our research coverage, and we plan to dig in further.

AI inference valuations are moving quickly, and we would build exposure across the group rather than pick a single name. fal, which runs inference for video, images and audio, is raising at a $16 billion valuation, double its $8 billion mark in April. Baseten is reportedly raising at $30 billion in a quick follow-on after $26 billion, up from $13 billion in June, and Fireworks AI is reportedly in talks at $30 billion to $40 billion, more than 2x this year. Our framework is an overweight to AI infrastructure across three buckets, data centers, inference providers and semiconductors, held at roughly equal weight, and if we had to overweight one it would be the inference platforms. Within each bucket, entry point matters. Together AI is valued below $10 billion with revenue we see as on par with peers valued at $15 billion to $40 billion, which looks like the better entry to us. In chips, we would rather own Positron at around $6 billion to $7 billion than Etched at $20 billion to $30 billion or more before it has a product in market. A proven founder can earn some premium, but at some point the revenue has to tie to the valuation.

Revolut's $115 billion valuation looks healthy for the business, but we question how much upside is left at this size. The leading fintech companies are now large enough that their private marks no longer line up with incumbent valuations: SoFi has a market cap of about $21 billion, roughly a fifth of Revolut's valuation. We would need to review Revolut's revenue in detail, but we find it hard to see a revenue gap large enough to justify a 5x difference. A move to $300 billion or $400 billion would put Revolut in the conversation with JPMorgan, which has a market cap of about $900 billion. Its international approach may work, but we are unsure how much growth remains in the category from here.

Stripe trades at about $190 billion in the secondary market, and of the fintech giants it is the one we like most, because AI agents could open a new market for its payments business. As agents start buying across the internet, they will need a provider to facilitate those transactions, and Stripe is building its own rails, including a blockchain system for tokenized payments, although regulation of tokenized payments is still unsettled. In that scenario Stripe could take Visa and Mastercard-like share in agent payments and become a company worth several hundred billion dollars. Even so, a move from $190 billion to $500 billion is a 2.5x return, and we think the largest private companies such as Databricks, Stripe and Revolut may be able to 3x over 10 years but not over three. Investors targeting 10% to 30% a year can get that in public stocks with daily liquidity, so for 2.5x to 4x-plus potential we look past the largest names to the next tranche: companies valued at $5 billion to $30 billion, such as Baseten, Together AI and Positron, that are not on the front page every day. With Anthropic's IPO roadshow underway and OpenAI possibly following next year, we expect those exits to make room for that next group.