Scale-Up InsightsAlphaGenerated June 12, 2026 · v1.0

What will it take to scale long-duration grid batteries (4-100 hours)?

Long-duration batteries can store renewable power for hours or even days, which a grid built on wind and solar needs to stay reliable. The leading designs use very different materials and factories. This report sets out the opportunity, then shows why the real test is turning low-cost, abundant materials into reliable systems at volume.

4

companies compared

6

material inputs

6

chokepoints

3

risk axes

The opportunity

What reaching volume would mean

Long-duration storage is what a grid running on wind and solar needs to stay reliable. Here is the opportunity across the market, the product, and the factory ramp.

Market

85-140 TWh by 2040

Reaching a net-zero grid could require 85 to 140 TWh of long-duration storage worldwide by 2040, a multi-trillion-dollar build-out.

Performance

100h at ~1/10 the cost

These batteries can store energy for up to four days at a time, something today's standard batteries cannot do affordably.

Timing

Factories ramping now

All four companies in this comparison are mid-way through building their first automated factories. The ones that industrialize first will set the cost curve for the sector.

Climate

Long-duration storage lets a grid rely on wind and solar continuously, displacing fossil-fired peaker and baseload plants.

Explore models on Koi

The bottom line

The active materials are low-cost and abundant, so the real test is industrialization: building first-of-kind factories and automated lines. Even Eos lost $44M in a quarter of record output.

Detailed findings

The full company comparison

Get the 6-input supply-chain map, all 4 company profiles, and the risk matrix across 3 axes, with the agents' reasoning and comparative takeaways.

See the reasoning behind each rating

See the agents' evidence-backed reasoning and comparative takeaways. Access every comparison with a single request.

Implications

What this means for you

The same findings, read for your decision.

Allocators

Spreading capital across the whole space

The opportunity is large, and execution decides who captures it: materials are not the differentiator here. Favor the companies furthest up the manufacturing learning curve, and remember that several still post losses while they scale.

Investors

Backing one company

Follow the factory, not the chemistry: yield, line uptime, and the path to positive gross margin. Eos's $44M quarterly gross loss on record revenue shows how far volume can run ahead of profitability.

Policymakers

Shaping incentives and supply security

The opportunity is a domestic manufacturing base, not scarce minerals, apart from bromine and vanadium, which are geographically concentrated. Support for first-of-kind factories and qualified electrolyte supply matters most.

Operators

Building or buying the technology

For multi-day storage, weigh the land and equipment each design needs. Form's iron-air needs large footprints, while flow systems add tanks and pumps. Match the design to your site, duration, and grid connection.

Turn this into a strategy

Rho's team works with stakeholders across climate finance, policy, and operations to turn this intelligence into action.

How this was made

Adversarially validated by AI agents

Six agents research, challenge, and refine every claim. Weak claims get removed before you see them.

Workflow orchestrator

Evidence validation cycle

Generate

Critique

Refine

Agent

01

Scout

Maps the company and picks the product line that matters.

Agent

02

Researcher

Traces critical materials and supply chains from primary sources.

Agent

03

Critic

Adversarially challenges every key claim.

Agent

04

Refiner

Removes anything the Critic disproved. Never invents.

Agent

05

Risk Analyst

Scores forward supply risks and scale-up failure modes.

Agent

06

Comparison Critic

Withdraws competitor comparisons that fail consistency checks.

Disclaimer: experimental alpha

Every finding on this page was generated by Koi's multi-agent AI research engine from publicly available sources. This analysis draws on public information only. The companies named were not consulted for it, and none has reviewed or endorsed it.

AI-generated research can contain errors, omissions, or outdated information. Severity ratings reflect agent judgment based on cited public evidence, not professional due diligence. Absence of a rating is not absence of risk. Nothing on this page is investment advice.

Are you a company named here? Request a review or correction.

Want scale-up analysis for your portfolio?
Join the alpha preview.