When a manager has compounded for 44 years with only five down years — the worst being a single
-10% — the question isn’t whether to listen. It’s what to actually take away. At Global Alts New York
2026, Seth Klarman, founder and CEO of Baupost Group, walked through how his firm is positioned for
what he called a market with “characteristics of a bubble” — without retreating to cash or capitulating to
the AI trade.
The most useful thing Klarman did on stage wasn’t predict a top. It was articulate the discipline that
allows a value-driven firm to underwrite a market this uncertain.
Bubble Characteristics — Without the Easy Call
Klarman did not call AI a bubble. He called the environment bubble-adjacent. “It has characteristics of a
bubble,” he said. “Optimistic tone around a transformational technology. New-era thinking.” His tell:
Allbirds, a shoe company, added “AI” to its name and the stock went up.
But he was equally careful on the other side. “AI seems like a technology that could be so
game-changing that it would be hard to dismiss it or call anything in particular around it a bubble.” The
hard part isn’t the technology — it’s the multiples. Winner-take-all or not? Which companies actually
win? At 40x or infinite multiples, investors are being asked to underwrite a very distant future. “None of
us can know with any confidence.”
Direct AI Exposure: ~10%, at Sensible Prices
Roughly 10% of Baupost’s book is in companies that directly benefit from the AI rollout — but only at
multiples Klarman is willing to defend. Amazon and Alphabet were the named examples: AWS
exposure at the former, internally designed chips plus AI surface area at the latter.
His operating principle around entry timing: “You don’t have to buy them every day. There are blips in
almost every company where you get windows of opportunity.”
The Optionality Play: Raw Land Adjacent to Power
The position that may travel furthest from this session is Baupost’s investment in raw land adjacent to
power infrastructure — sites that could become data centers. Crucially, Baupost is not planning to
build. The two criteria Klarman named were political permitting and access to power; the structure is
pure optionality at the land basis.
Internationally, the firm took a private position in non-China Asia data centers, acquired via a spinout at
roughly a 40% discount to public market multiples. Same theme, very different entry point.
What AI Won’t Touch — The Quiet Long Book
This was the section that should resonate with allocators thinking about long-tail equity exposure inside
their alternatives sleeves. Klarman described what he called “AI agnostic” businesses: roofing, housing
supplies, travel infrastructure — companies whose end-demand and operating model are not at
meaningful risk from large language models. “People aren’t paying attention,” he said, “and the price is
drifting lower.”
The mirror trade is in credit. Klarman is looking at perceived “AI losers” — software-related credits
getting clobbered, trading at very low cash flow multiples. “At least something to take a look at.”
Private Credit Opportunities 2026 — Idiosyncratic, Not Systemic
Klarman is not calling a credit cycle. But he is seeing more idiosyncratic distress — a Brazilian
corporate restructuring, a large PE deal with impaired debt and an exchange offer that has left
mispriced securities behind. His framing of why credit dislocates well: “Credit is inherently interesting
because of the transition of ownership. When bonds get downgraded, people dump. When they file for
bankruptcy, people dump. Even distressed funds eventually move on — and new capital is needed.”
He believes the market is “due for a credit cycle” without forecasting the timing. For LPs underwriting
private credit opportunities 2026, that’s a useful distinction — between cyclical positioning and
idiosyncratic deployment.
Commercial Real Estate Distressed — The Top Conviction
The single highest-conviction idea Klarman shared was assisted living. Post-Covid, many newly built
facilities couldn’t reach occupancy and a wave of bankruptcies followed. The system, he said, is
starting to clear.
His broader commercial real estate distressed framing was equally direct. “The logjam is starting to
break. Fundamentals are starting to improve. We’re seeing opportunities to deploy capital at significant
discounts to replacement cost with very attractive returns — no heroic assumptions.” Industrial land,
warehouses, and cold storage — all driven by the onshoring trend — were called out as areas of
continued demand. His structural advantage: “I love being below the radar of the big gorilla firms.”
The Macro Overlay
Klarman did not hide his concern on US debt — 100% of GDP, $2T+ structural deficits, a path to $50T
in five years. He flagged the Strait of Hormuz as underpriced, with storage drawn down quickly and oil
potentially at $150+ if it closes for months. He framed the AI build-out as a growing political and NIMBY
risk at the local level. On the Fed: he expects patience — one or two hikes are possible, but a cut is the
preferred path.
The Takeaway
Klarman’s framework in one line: in markets this uncertain, the work is to own what you can underwrite,
leave optionality where the upside is asymmetric, and let the rest of the market provide the entries.
Value investing alternatives AI era isn’t a style box — it’s a discipline about what you’ll pay for
confidence.
Sessions like this one are why allocators continue to use iConnections to find managers operating
with this kind of clarity. Global Alts New York 2026 brought together the firms underwriting the same
market from very different angles — Klarman on the long side of what AI can’t touch, Chanos on the
short side of what’s being overpaid for around it