VS. NOTHINGBY HERITAGE HOLDINGS

INVESTIGATION · SOURCE PUBLISHED JULY 2026

NO TRADE DECLARED

Money can buy the chips.
It cannot summon the power.

Ribbit Capital says the AI bottleneck is moving from money to electrons. We are tracking the harder question: who can deliver usable power in time - and is acting on that thesis really better than doing nothing?

SOURCE THESIS RIBBIT CAPITAL · POWER LETTER · JULY 2026 OPEN RIBBIT'S PUBLIC SOURCE
EVIDENCE MONITORING NO MONEY GRADE NOT A LIVE FIGHT LAST CHECKED
01

CAPTURE THE CALL

What did Ribbit actually say?

"The bottleneck today ... is the flow of electrons."

RIBBIT CAPITAL · POWER LETTER · JULY 2026

Ribbit's 2025 Token Letter argued that cheaper, more abundant digital intelligence would make rare physical capabilities more valuable. Its 2026 Power Letter identifies the next constraint: the infrastructure that creates electrons and turns them into intelligence.

Ribbit maps both stacks through Make, Move, Store, and Sell. VS. Nothing narrows that broad venture thesis into a claim evidence can attack.

THE TESTABLE VERSION

The scarce asset is not energy in the abstract. It is firm electricity that can be connected, conditioned, and delivered at the right location by the date a compute campus needs it.

THE THESIS IS NOTBuy energy.
THE THESIS ISOwn the ability to deliver and use power on time.
02

IS THE CALL REALLY BETTER THAN DOING NOTHING?

Not decided. No trade was declared.

RIBBIT'S THESIS POWER IS THE BOTTLENECK Institutional thesis · source dated July 2026
DOING NOTHING NOT DECLARED No benchmark, basis, or hold treatment was specified
NO MONEY RESULT · INVESTIGATION ONLY

The letter does not specify a public instrument, direction, starting value, or fixed horizon. We will not invent those after the fact. Every power expression below remains a separate creator Call and can become a Fight only after its own starting conditions and doing-Nothing comparison are locked.

03

THE POWER MAP

One constraint. Five very different exposures.

MAKE

Create firm electrons.

Generation, on-site power, fuel, advanced geothermal and nuclear.

TRAP · The technology may arrive after the shortage.
MOVE

Connect and carry them.

Transmission, transformers, switchgear, substations and grid engineering.

TRAP · Backlog can be cyclical or regulated.
STORE

Move power through time.

Grid batteries, thermal storage, long-duration storage and orchestration.

TRAP · Cheaper hardware does not guarantee project returns.
SELL

Price and deliver flexibility.

Retail power, virtual power plants, tolling, aggregation and risk markets.

TRAP · Regulation may own the customer.
CONVERT

Turn electricity into intelligence.

Power management, cooling, memory, networking and compute factories.

TRAP · Hardware cycles shorten and customers internalize capacity.

CONVERT is our reading aid, not a fifth Ribbit bucket. It makes visible the bridge between Ribbit's electron-side and compute-side versions of Make, Move, Store, and Sell.

04

CHALLENGE THE THESIS

Four things must survive contact with reality.

DEMAND

Usage outruns efficiency.

AI and data-center electricity demand must grow faster than models and chips reduce the electricity needed for each unit of intelligence.

DELIVERY

The grid stays slower than compute.

Generation, permits, transmission, transformers, substations and interconnection must remain unable to expand on software time.

ECONOMICS

The bottleneck owner keeps the premium.

Customers must pay for earlier, firmer power - and suppliers must convert that urgency into margins and cash rather than just revenue.

VALUATION

The right story is not already overpaid.

A security can lose even if the power thesis is right. Leverage, dilution, execution, and the starting multiple still count.

05

REALITY GRADES

The mechanism is real. The outcome is not settled.

These sources support a demand-and-delivery constraint. They do not prove that any particular security is cheap or will outperform. The latest LBNL figure also supersedes the Power Letter's dated 2.6 TW queue figure for today's evidence layer.

NEXT WATCHED
  • EIA load forecasts and realized demand
  • LBNL queue size, withdrawals, and completion time
  • Firm offtake, energization dates, and in-service megawatts
  • Backlog conversion, margins, cash flow, cancellation, and dilution
06

ONE GRID. SEPARATE CALLS.

Agreement on power does not make these the same bet.

These are independent creator Calls already captured on VS. Nothing. They are not Ribbit recommendations, not components of a hindsight basket, and not yet sealed Fights.