Edgen
Try Ed Free
Content
What Aschenbrenner Is Actually Buying
About That $1.6B NVDA Put — Why It's Probably a Hedge
What the Other 4 Big-Name Investors Are Building
How to Build AI Wealth Without Chasing One Stock
Track Smart Money Moves — Don't Copy Them

Markets Confusing? Ask Edgen Search.

Instant answers, zero BS, and trading decisions your future self will thank you for.

Try Search Now

Leopold Aschenbrenner's Real AI Bets: Bloom Energy, SanDisk, CoreWeave — the NVDA Put Is Probably a Hedge

Edgen
· May 21 2026
Leopold Aschenbrenner's Real AI Bets: Bloom Energy, SanDisk, CoreWeave — the NVDA Put Is Probably a Hedge

By Edgen Research | 2026-05-21 Tickers covered: $BE, $SNDK, $CRWV, $GOOGL, $MSFT, $AVGO, $NVDA | Landscape: $BRK.B, $AMD, $MU, $TSM

What are the world's most-watched investors actually buying right now? The Q1 2026 13F filings — released May 15 — give the clearest answer in years: every major filer is building AI exposure. The disagreement is over how.

The most-headlined case is Leopold Aschenbrenner — the ex-OpenAI researcher who wrote the Situational Awareness essay and now runs Situational Awareness LP ($9.28B regulatory AUM per the March 2026 SEC Form ADV; $5.52B of that is the disclosed U.S. equity book). The press ran with "Aschenbrenner bet $8.46 billion against AI chips." That headline is technically true and practically misleading. What he actually did with real capital: load up on the AI infrastructure that makes the boom possible — power, memory, data-center capacity. The $8.46B against NVDA and other chipmakers is almost certainly a hedge on those AI longs, not a directional bet against AI itself.

If you're trying to build AI exposure for the long run, his buy list — and what 4 other big names are doing alongside it — is the more useful map.

What Aschenbrenner Is Actually Buying

Behind the dramatic short headline is a portfolio that's overwhelmingly long the AI infrastructure stack:

Position Size What he's building exposure to
Bloom Energy (BE) $878M On-site power generation — the bottleneck for new AI data centers
SanDisk (SNDK) $724M NAND memory — the unsung half of every AI chip system
CoreWeave (CRWV) $556M AI GPU rental — direct revenue from AI compute usage
Crypto miners (CLSK / RIOT / APLD / IREN) Added more Existing data centers + power contracts pivoting to AI hosting
Other equity longs (disclosed in 13F) ~$3.4B more His total equity book sits around $5.52B

The pattern: he's owning what gets consumed as AI scales, regardless of whether NVIDIA's stock price is right at this exact moment. In his own words, from the 2024 essay that launched the fund:

"Over the past year, the talk of the town has shifted from $10 billion compute clusters to $100 billion clusters to trillion-dollar clusters. Every six months another zero is added to the boardroom plans." — Leopold Aschenbrenner, Situational Awareness: The Decade Ahead (June 2024)

If you believe compute really does scale that way (Aschenbrenner does — and his $9.28B fund is now built around that belief), then the names that sell the inputs win regardless of which chip company captures the most surplus.

Aschenbrenner's AI Long Book — 3 retail-buyable channels: Bloom Energy 41%, SanDisk 34%, CoreWeave 26%
Source: SEC Form 13F-HR, Situational Awareness LP, filed May 18, 2026 — Aschenbrenner's three disclosed AI infrastructure positions, equity values at market.

You can buy any one of these names in a regular retail account. You can't easily replicate a hedge fund's $1.6B options position, but the long book is a 3-channel map any retail investor can follow.

About That $1.6B NVDA Put — Why It's Probably a Hedge

This is where most of the press got the story wrong. The mechanics of 13F filings make options positions look much bigger than they actually cost.

What 13F shows What it actually means
"$1.6B notional put on NVDA" The value of stock the puts could be exercised against — not what he paid
"$8.46B notional chipmaker put basket" Same — total stock value of the puts' underlyings
What you can't see: premium actually paid Industry typical: OTM puts cost roughly 2–5% of notional, so $8.46B notional likely represents ~$170-420M in real cash premium

There's also a structural giveaway: the fund's $9.28B regulatory AUM cannot cash-collateralize $8.46B of puts. That math only works if these are options paying premium for asymmetric exposure — i.e. insurance, not a fully-funded directional short. Applied to NVDA specifically: a $1.6B notional put might represent ~$30-80M in actual premium. That's standard hedge sizing against a $5.52B AI-tilted long book. The puts protect the longs.

Why '$8.46B short' is misleading — notional vs actual premium paid: full chipmaker put basket $8.46B notional vs ~$170-420M premium; NVDA puts $1.6B notional vs ~$30-80M premium
Aschenbrenner's chipmaker put basket as notional vs estimated actual premium. OTM puts typically cost 2-5% of notional. The fund's $9.28B regulatory AUM cannot cash-collateralize $8.46B of puts — supporting the hedge interpretation.

Compare that to:

  • A real $1.6B bet against NVDA would show up as $1.6B in cash deployed — equivalent to outright shorting the stock. That's not what he did.
  • What he did do: small-premium puts that pay off big if there's a chipmaker pullback, otherwise expire near-worthless. Classic AI-bull hedge.

The financial press is divided on this — Hedge Fund Alpha called it a "$8.5B bet against AI chips," while TrendSpider more carefully framed it as "semi put notional while holding AI infrastructure longs." We side with TrendSpider's read. The dollars deployed in the long book are the actual conviction.

What the Other 4 Big-Name Investors Are Building

Stripping the same lens onto the rest of the Q1 13F season — what's the actual long-term exposure each filer is buying?

Investor What they're building exposure to In their own voice
Greg Abel (Berkshire — Buffett's successor, first 13F as CEO) Quadrupled Google (+204% to ~$23B); Delta Air Lines; NYT Buffett & Munger spent years publicly calling Google "perhaps our biggest missed opportunity." Abel's +204% closes that gap — AI exposure through ad cashflow, not chip cycles
Bill Ackman (Pershing Square) NEW Microsoft $2.09B position; more Amazon; cut Google 95% On X May 15: "Microsoft's 365 + Azure are two of the most valuable franchises in enterprise technology" — bought at 21× forward P/E, in line with market multiple
Nancy Pelosi (Congressional disclosure) Deep ITM call options on NVDA + Broadcom Her PTR pattern: deep-in-the-money calls = stock-like upside for ~30% upfront capital. High-leverage long-term AI bull
Donald Trump (Presidential OGE disclosure) NVDA, MSFT, AVGO, AMZN, AAPL ($1–5M each); big sells of AMZN, META, MSFT Trump Org statement: "holdings maintained exclusively through fully discretionary accounts independently managed by third-party institutions." 3,642 individual trades in 90 days reads otherwise

See every big-name investor's full holdings — Edgen Whale Holdings →

How to Build AI Wealth Without Chasing One Stock

The lesson stacked across all six filers: smart money is building AI exposure through many different channels, not just chips. If you're trying to participate in AI for the next 5-10 years, you don't have to bet everything on NVDA's next earnings.

5 channels to build AI exposure: Software margins (MSFT, GOOGL); Power generation (BE); Memory & storage (SNDK, MU); Compute rental (CRWV + miners); Chip leaders (NVDA, AVGO — highest beta, valuation risk)
Five channels smart money is using to build AI exposure across Q1 2026 13F filings. Each channel maps to a retail-buyable ticker set.

You don't have to pick one. Most of the smart-money filers above own combinations across these channels — that's how a real AI allocation looks in a long-term wealth book. Want deeper coverage on the memory side? Our SK Hynix / Micron piece digs into how the HBM cycle is repricing the memory channel right now.

Track Smart Money Moves — Don't Copy Them

13F filings arrive 45 days behind the actual trades. By the time you copy Pelosi, her timing is already six weeks stale. By the time you act on a 13F disclosure, the position may already have reversed.

The smarter use is to watch the pattern of where smart money keeps showing up. Edgen Whale Holdings tracks every SEC filing the moment it hits — that's the closest a retail account gets to seeing where the real money flows.

This quarter's pattern is clear: power, memory, software margins and data-center capacity. Aschenbrenner is building inside all four. The NVDA put is just the insurance behind it.

Research and education, not personalized investment advice. Primary sources: SEC Form 13F-HR filings for Berkshire Hathaway, Pershing Square Capital Management, ARK Investment Management, and Situational Awareness LP (filed May 15–18, 2026); U.S. House Clerk PTR for Nancy Pelosi; OGE Form 278-T for Donald J. Trump (certified May 8, received May 12, 2026). Secondary verification via Quiver Quantitative, TrendSpider, Hedge Fund Alpha, Seeking Alpha, WhaleWisdom, CNBC, Benzinga. Verify against current data before acting. Edgen and contributors may hold positions in securities discussed.

Recommend
Nomura raised SK Hynix's price target from ₩840K to ₩4M in 6 months — 5x. Goldman lifted MU's 2026 EPS estimate 133%. Why MU and SNDK move from Hold to Buy.

Nomura Quintupled SK Hynix's Target in 6 Months — A Tailwind for MU and SNDK

On May 17, Nomura raised SK Hynix's price target to ₩4 million won — roughly five times the level it set just six months ago. The size of that revision matters less than what Nomura wrote underneath it: AI memory demand is entering a "new regime" that breaks the historical cycle model. We're updating our #66 AI hardware basket map on the back of it — MU and SNDK move from Hold to Buy. The trajectory is the story, not the endpoint. Four sequential raises. Each one looked aggressive when published. Each one was conservative within weeks. The current PT implies +119.9% upside from the May 15 close of ₩1.82M — Nomura's bull case explicitly recalibrating to a "new regime," not extrapolating from the prior cycle. I'd argue the May 17 raise matters more than the four that came before it. The earlier raises tracked rising EPS estimates inside the same cyclical model. This one reframes the model itself — what Nomura calls "memory stock re-rating driven by exponential demand growth from AI." Nom
Edgen
·
May 18 2026
NVDA isn't the whole AI hardware trade. SanDisk +3,362%, Micron +340% past year prove it. Which AI hardware sectors still have upside — and which don't?

NVDA Is One Basket of AI Hardware. Here Are the Other Four, Ranked.

SanDisk is up 3,362% over the past 12 months (Yahoo Finance). Micron is up around 340% in the same window (Yahoo Finance). Both are AI hardware. Neither is NVIDIA. If you call yourself "long AI" but own only NVDA, you're not long AI — you're long one basket, and the other four ran without you. SNDK and MU are the proof of how expensive that miss can be. Here's the map. Five baskets. Where I'd put new money today (Basket 4) and where I wouldn't (Basket 1). Bottom line first. If you can only add to one basket today, I'd add to Basket 4 — Optical Interconnect. Not NVDA. Not HBM. Not foundry. One sentence: AI's bottleneck is shifting from "memory can't feed the GPU fast enough" (the 2024 problem) to "the wires can't connect 100,000 GPUs at once" (the 2026-2028 problem). The market hasn't fully priced this yet. AI-themed funds barely own any of the optical names.
Edgen
·
May 15 2026
earnings-beats-fail-2026-edgen-3-signal-framework_header.png

Why Earnings Beats Fail in 2026: Edgen's 3-Signal Framework

Something broke in Q1 2026 earnings that contradicts what most retail investors learned in 2023-2024. Affirm reported adjusted EPS of $0.37 against a $0.27 consensus — nearly 40% above the Street — and the stock fell 5%. Coinbase missed revenue by 31% — a catastrophic shortfall by any historical standard — and the stock declined only 2.5%. SoFi beat the quarter and raised full-year guidance, then dropped 15%. Palantir reported 85% year-over-year revenue growth — and went sideways. These are the four prints Edgen Research covered in real time during April and May 2026. Across them, the relationship between an earnings beat and stock direction has visibly decoupled. This is not a noisy small sample — it is a structural change in how the market processes quarterly information. Three signals appeared in every single beat-and-fade case we tracked, and they also predict — in reverse — the rare cases that beat and rally, such as NXP's +18% reaction to its Q1 print. This article is the framewo
Edgen
·
May 14 2026
affirm-investor-forum-may-12-2026-4-numbers-decide-stock_header.png

Affirm Investor Forum May 12: 4 Numbers That Decide the Stock

Affirm crushed Q3 — adjusted EPS $0.37 against a $0.27 consensus, revenue $1.12B against $1.06B expected, FY26 guide raised to $4.18B-$4.21B. The stock closed Friday at $64.01, down 5%, after sitting flat through the after-hours session. Same thing happened on Q2: beat, then dipped 4.41%. Two consecutive beat-and-fade reactions tell you what the market is actually waiting for, and that thing arrives Tuesday, May 12, 2 PM to 5 PM Eastern, in person in New York and live on Affirm's IR webcast. The forum delivers something Affirm has not given investors since the 2021 IPO: a medium-term financial framework. Five years of "here is the next quarter and a vague long-term ambition" — finally getting structure. Four specific numbers will land on those slides, and those four numbers decide which of three scenarios is playing out at Wednesday's open. They also tell you whether the sell-side's $75-90 zone (Morgan Stanley $79, Needham $90, Oppenheimer $87) is the right anchor, or whether our $95 B
·
May 11 2026
Coinbase missed Q1 by $70M and posted a $394M GAAP loss — yet the stock dropped just 2.5%. Why the bear case isn't paying out, and the OCC trade still works.

Coinbase Missed Q1 by $70M — Why the Trade Hasn't Broken

Coinbase reported Q1 2026 after the close on Wednesday and the headline was ugly: revenue $1.41 billion versus the $1.48 billion sell-side number, a 31% drop year-over-year. GAAP net loss of $394 million, or minus $1.49 per share, dragged down by $482 million in unrealized losses on the Bitcoin sitting on the balance sheet. Transaction revenue $755.8 million and subscription and services $583.5 million both came in light. And the stock barely moved. COIN closed Wednesday at $192.96, down 2.53% on the day, and gave back another 4% in after-hours. That's not the reaction tape you'd expect on a 31% revenue miss. It's the reaction tape of a stock where the bad news was already in the price. Three numbers nobody put in the headline tell the actual story. Adjusted EBITDA $303 million — the thirteenth straight positive quarter even with the print falling apart. Trading volume market share at an all-time high of 8.6%. Stablecoin revenue $305 million, up 11% year-over-year on USDC growth. Coinb
·
May 08 2026
affirm-q3-2026-beat-raised-guide-may-12-forum-thesis_header.png

Affirm Q3 Crushed Estimates. Why Didn't the Stock Move?

Affirm just posted one of the cleanest fintech prints of the quarter. Revenue came in at $1.039 billion versus the $1.012 billion consensus — a +27% year-over-year line. Adjusted EPS hit $0.30 against a $0.17 estimate, nearly double the number Wall Street was modeling. GMV ran $11.6 billion (+35% YoY), the biggest quarter Affirm has ever processed. Management raised the FY26 revenue guide to $4.175–4.205 billion (up from $4.086–4.146B) and lifted the Q4 range to $1.08–1.11 billion. Then the stock did nothing. AFRM popped roughly 2% after-hours on the headlines, faded through the Thursday session, and closed at $67.50, down 0.21%. A beat-and-raise this clean usually moves a stock 5–10%. This one moved less than a quarter of a percent. That non-reaction is the story. We flagged this exact risk in our May 1 piece on the May 12 investor forum — the print itself was likely already priced, and the real binary event is four days away. Q3 was the warm-up. Here's why the tape behaved exactly th
·
May 08 2026
ZEC just hit $590 in 3 days — nearly our $600 PT from Monday. Now what? Monero (XMR) at $409 hasn't moved. Here's the privacy coin rotation play.

Monero vs Zcash 2026 — Which Privacy Coin Should You Buy?

Monero (XMR) trades around $409 going into a privacy upgrade audit nobody is talking about. From May 11 to May 22, a team of cryptographers will tear apart Monero's biggest protocol change in years — FCMP++, short for Full-Chain Membership Proofs. If it ships clean, Monero stops relying on the small "ring" of decoy senders it has used since 2017 and starts hiding every spender behind the entire chain history. That's a real privacy upgrade, not a marketing one. While that audit runs, Monero already does the thing every retail investor thinks Zcash does: every transaction is private by default. No shielded address, no opt-in. About 150,000 transactions a day, 100% of them hidden. Mining runs on a CPU-friendly algorithm called RandomX, so the network isn't dominated by a few industrial farms. And here's the catch most coverage glosses over — you can't easily buy XMR in the United States anymore. Coinbase, Binance.US, and Kraken (in several US jurisdictions) have all delisted it. Here's th
·
May 06 2026
zcash-zec-stock-400-surge-2026-privacy-coin-narrative-bitcoin-younger-sibling_header.png

Why Zcash Just Broke $400 — Bitcoin’s Younger Sibling

If you watched crypto Twitter on Sunday, you saw the same chart everyone else saw. Zcash (ZEC) — a coin most retail had written off as a 2017 leftover — ripped from the low $300s to $424 in a few sessions. First time above $400 since January. Network value back over $7 billion. About $10.5 million of short positions blown up on the way. It wasn't a meme pump. The week before, Grayscale's Zcash Trust did roughly double its usual volume. The share of ZEC sitting in private, on-chain "shielded" wallets hit an all-time high of 30%. And two of the loudest voices in macro-crypto — Raoul Pal and Barry Silbert — said the same thing in different words: privacy is the next thing the market wants, and Zcash is the cleanest way to own it. Our call on Zcash: Buy, $600 price target. That's roughly 41% above where it trades today. The case rests on three layers — the narrative is real, the on-chain data backs it, and a structural quirk in the privacy-coin market makes ZEC, not its more famous rival M
·
May 05 2026
Biren 06082.HK opened +76% in January and has barely moved in four months. The cornerstone lockup is two months out. So is the better Chinese chip bet.

Biren (06082.HK) Stock 4 Months After IPO: GPU vs Photonics Bet

Four months ago, Biren Technology — Hong Kong ticker 06082.HK — opened its first trading day at HK$35.7 against an IPO price of HK$19.60 and closed up 76%. Retail subscribed the deal 2,300 times over. It was the first Chinese GPU startup ever to list in Hong Kong, and the headlines treated it like a NVIDIA-of-China moment. Today the stock sits around HK$36 — basically where it closed on debut. The mania didn't keep going. It also didn't fall apart. What it did was settle, and now we're three months from the first cornerstone lockup and one trading session away from a question almost nobody is asking out loud: does owning Biren still make sense, or is the better Chinese AI chip bet sitting one ticker over at 1879.HK? That second ticker is Lightelligence (1879.HK) — the silicon-photonics company that opened on April 28 and closed +383% in one session. Same week as Biren's quiet 4-month checkpoint. Same end market. Same theme — Chinese AI hardware self-sufficiency. Different physics. We c
·
May 05 2026
Q1 revenue is going to fall 26%. But while everyone watched the print, Coinbase quietly got something no crypto firm has ever had.

Should I Buy Coinbase Before May 7? The OCC Trust Bank Bet

Most people staring at Coinbase right now are looking at the wrong screen. The screen everyone's watching is Thursday's Q1 print — May 7, after the close. And that one is going to be ugly. Bitcoin fell 22% in the quarter. Ether fell 41%. Global exchange volume is down nearly half from its October peak. Wall Street already wrote the headline: "Coinbase revenue plunges 26% as crypto winter returns." That story is already in the price — COIN sits around $220, basically flat for the year while the S&P grinds higher. The screen nobody's watching is the one that landed two weeks ago, when the Office of the Comptroller of the Currency — the federal banking regulator — gave Coinbase conditional approval to operate a national trust bank. First crypto-native company to ever get one. That's not a 90-day story. It's a structural unlock that shows up in models years from now, and Wall Street hasn't moved its targets yet. So the question retail is actually asking is the right one: should I buy C
·
May 05 2026

Your money person, finally.

Try Ed free. No credit card. No commitment.
Try Ed Free