AI Funding Portfolio
Disclosed AI corporate equity stakes, compute commitments, and capacity partnerships, each tagged with the recipient silicon dependency.
Generic funding trackers (including our own /funding rounds tracker) list deal size and date. This registry adds the silicon side: every entry pairs the financial flow with the AI hardware the recipient actually runs on. That is the lens we used in the Nvidia $40B equity book article to argue that the 2026 Nvidia portfolio is best read as defense, not growth: a $30B option on OpenAI staying primarily on Nvidia, $3.2B on Corning supplying optical interconnect for the next two Nvidia generations, $2.1B on IREN deploying 5 GW of Nvidia DSX-branded racks.
The most useful single number for tracking whether the loop is working is the share of frontier training compute (in FLOPs) that runs on Nvidia in any given quarter. We will start publishing that derived metric once enough quarters of disclosure have rolled in.
Free agent endpoint
/api/funding/portfolioFull registry plus summary aggregates by silicon dependency, deal type, and investor. Filters:?silicon_dependency=&type=&from=&to=&since=&until=. Each entry returns id, from, to, amount_usd_max, amount_usd_disclosed, announced_date, type, recipient_silicon_dependency, commercial_quid_pro_quo, source_urls, notes.
Frequently asked questions
- What does this registry cover?
- Disclosed AI capital commitments where the source is a SEC filing, hyperscaler press release, or reputable trade reporting. Equity stakes, compute purchase commitments, and capacity partnerships all qualify. Each entry is tagged with the recipient's primary silicon dependency at announcement (Nvidia, TPU, Trainium, MI400, Maia, or mixed) so an agent can analyze the customer-investor loop directly.
- How is this different from /funding (the rounds tracker)?
- /funding tracks AI startup financing rounds (seed, Series A through E, growth) with stage and post-money valuation. /funding/portfolio tracks corporate AI equity stakes BETWEEN existing players (Nvidia investing in OpenAI, Google in Anthropic, etc.) plus compute commitments and capacity partnerships, tagged by recipient silicon dependency. Different lens, complementary registry.
- Why the silicon-dependency tag?
- The customer-investor loop only makes sense when you can see the silicon side. A $30B equity stake from a chip vendor in a frontier lab pencils very differently when the lab runs primarily on that vendor's silicon (defense) versus on a competitor's (a hedge). Tagging the silicon dependency explicitly lets the registry serve as input to that analysis rather than just a deal log.
- Who curates the registry?
- TensorFeed editorial. The underlying public-record sources retain their original license; the TF curation, tagging, and commercial_quid_pro_quo summaries are CC-BY 4.0. Each entry preserves the canonical source URL so end-users can verify.
- How fresh is it?
- Editorial cadence. An entry is added when a related original article goes out, or when reporting picks up something we have not covered. Smaller and more carefully curated than automated CB scraping; we would rather ship 10 well-sourced entries than 100 with stale tags.