Rob's Portfolio Advisory Board

Stance

Yellow. A thoughtful warning on AI hardware economics, not a call on any holding.

Signals timelinenewest first · dashed = secondary / unverified

Oct 19 — quiet; AI private-credit charts

No Chamath posts. Social Capital kept charting AI-buildout financing; All-In ran a Figma interview.

On Oct 19, All-In published an interview with Figma CEO Dylan Field about the “SaaSpocalypse.” Source: X
On Oct 19, Social Capital posted that commercial power use in Virginia grew about 7x the US rate, mainly because of data centers. Source: X
On Oct 19, Social Capital posted that tech companies' share of direct loans rose from 22% in 2010 to 44% in 2025. Source: X

Oct 16 — CoreWeave debt / AI-trade crack

All-In led with the CoreWeave debt selloff; Social Capital quantified CoreWeave's debt load and re-promoted its private-credit Deep Dive.

On Oct 16, Social Capital posted that annual US data-center construction spending is about 1.9x spending on office buildings. Source: X
On Oct 16, Social Capital shared its Deep Dive asking whether the AI buildout is leading to a private credit bubble. Source: X
On Oct 16, Social Capital posted that CoreWeave's net debt grew from $6.7B to $29.8B in 18 months, with interest at about 25% of revenue. Source: X
On Oct 16, The All-In podcast asked whether the AI trade just cracked after CoreWeave tanked in a debt selloff. Source: X

Oct 15 — AI ARR quality + VC concentration

Chamath attacked AI ARR reporting abuse and AI-to-AI VC revenue quality; Social Capital charted AI ~83% of US VC funding.

On Oct 15, Social Capital posted that through September, AI companies captured about 83% of all US venture funding. Source: X
On Oct 15, Chamath said he suspects 70–80% of AI-company revenues come from other VC-funded AI firms, which he called a quality problem. Source: X
On Oct 15, Chamath said VCs enabled AI companies to bastardize revenue reporting by blurring Annualized Run Rate with contractual SaaS ARR. Source: X

Oct 14 — Anthropic IPO + hyperscaler debt

Chamath stuck with a ~$1T Anthropic clearing price; Social Capital posted hyperscaler debt and NVDA customer concentration.

On Oct 14, Social Capital posted that 44% of NVIDIA's revenue came from three direct customers in the first half of fiscal 2027. Source: X
On Oct 14, Social Capital posted that five hyperscalers added $164B of net new debt in the first half of 2026, about twice the amount for all of 2025. Source: X
On Oct 14, Chamath said Anthropic needed to get public at $200B while the market is clearing it at “+/- $1T,” and that his mindset is unchanged. Source: X

Key quotes & findings

On Oct 22, Chamath posted on X: “This SpaceX paper is money good and better than most sovereign debt.” Source
On Oct 21, Social Capital posted on X: “~46% of bonds sold this year to finance US data centers are junk-rated, ~2x the share in 2025.” Source
On Oct 20, Chamath posted on X: “As a result, the revenue quality in AI is currently poor.” Source
On Oct 20, Chamath posted on X: “We will need to transition from speculative AI spending to consistent, reliable spend that shows up in public market earnings reports.” Source
On Oct 19, Social Capital said: “Tech companies have increased their share of direct loans from 22% in 2010 to 44% in 2025.” Source
On Oct 16, Social Capital said: “CoreWeave’s net debt has grown from $6.7B to $29.8B in 18 months. Interest now consumes 25% of its revenue.” Source
On Oct 16, Social Capital said: “We put together a Deep Dive exploring whether the AI buildout is leading to a private credit bubble.” Source
On Oct 16, Social Capital said: “Annual construction spending on US data centers is now ~1.9x that of office buildings.” Source

On Oct 28, Chamath quoted Alex Song's post on GPU financing: 'Super interesting comparison of GPU finance and the AI buildout to airlines. I've seen two other people make this comparison now: Ken Griffin, Sridhar Ramaswamy. Both are very, very smart so worth learning more.' A reply said model efficiency makes GPUs different; Chamath answered that route planning might be a better analogy. Nothing new from All-In beyond an event promo.
No TV/print interview or trade disclosure in window. Only one Social Capital newsletter issue is in Gmail.

What to watch next

Watch for the next Social Capital research newsletter issue; the last one came out on Oct 10.
Watch Meta's Q3 results on Oct 21 for a read on AI spending.

All sources27 links · open in a new tab · dashed = secondary / unverified

Overview quick notes

On Oct 28, Chamath quoted Alex Song's post on GPU financing: 'Super interesting comparison of GPU finance and the AI buildout to airlines. I've seen two other people make this comparison now: Ken Griffin, Sridhar Ramaswamy. Both are very, very smart so worth learning more.' A reply said model efficiency makes GPUs different; Chamath answered that route planning might be a better analogy. Nothing new from All-In beyond an event promo.

Why (overview note)

If GPUs end up financed and priced like airplanes, returns on AI hardware could get squeezed, which matters for NVDA and for the big spenders like GOOGL and META.

Anthropic IPO pricing · next All-In episode · Nov 3 midterms

← Back to Board Update

Chamath Palihapitiya T2 🟡 WATCH

SIGNAL

TICKERS / BRANDS