The most honest way to think about a grid battery is as a financial instrument. It buys electricity when it's cheap, sells it when it's expensive, and the difference is its only real job. The problem is that the price of power in deregulated markets like Texas can swing by 300% in five minutes, a pace that makes human traders look like they're moving through molasses. Rewbi, a San Francisco startup founded last year, is betting that an AI can do it better. It has taken a $4 million seed round to prove it, and has already registered as a power company with ERCOT, the Texas grid operator [Rewbi, May 2025] [Y Combinator, Unknown].
The algorithmic wedge
Rewbi's model is straightforward. It rents battery capacity from asset owners for a fixed monthly fee. Its AI then takes over dispatch, deciding precisely when to charge and discharge to capture price spreads in the wholesale power market. The company claims this optimization earns more than double the fixed rental fee in monthly revenue, splitting the surplus with the battery owner [Y Combinator, August 2024]. The wedge is pure speed and data processing. The AI tracks hundreds of live inputs,from weather forecasts and grid congestion to natural gas prices,to make decisions faster than any human team could [Orange Collective, October 2024]. For a battery owner, it turns a complex, risky trading operation into a predictable SaaS-style revenue stream.
A founder built for this fight
This is not CEO Thomas Marge's first rodeo in energy storage. His previous company, inBalance, was acquired by Stem, a publicly traded leader in AI-driven energy storage systems [Orange Collective, October 2024]. That exit gives him a rare combination of domain credibility and a clear understanding of the incumbent landscape. Co-founder and CTO Derek Modzelewski brings a different kind of pedigree, having been an early technical hire at Adept, an AI agent company [Orange Collective, October 2024]. The team, still under ten people according to LinkedIn, is built to bridge the gap between high-stakes power markets and cutting-edge machine learning [LinkedIn, Unknown].
The early traction and the Texas beachhead
Rewbi's most concrete signal of progress is its registration with ERCOT. Becoming a Qualified Scheduling Entity in the Texas grid is a non-trivial regulatory and technical hurdle; it means Rewbi can directly participate in the market, bidding its aggregated battery capacity like a traditional power plant [No Cap Blog, Unknown]. This is the essential plumbing for its business model. On the financial side, the company closed a $4 million seed round in May 2025, following an earlier $500,000 pre-seed [Rewbi, May 2025] [Crunchbase, September 2024]. The investor list is a long tail of climate and tech funds, including Y Combinator, SurgePoint, General Catalyst, and Collab Fund.
The company's reported financial trajectory shows early revenue generation.
Pre-seed (Sep 2024) | 0.5 | M USD
Seed (May 2025) | 4 | M USD
Reported Revenue (2025) | 0.33 | M USD
Where the model could short-circuit
Optimizing batteries for arbitrage is a well-understood concept, which means Rewbi is not alone. Its most direct competitor appears to be Gridmatic, another AI-powered energy trader. The core risks for Rewbi are not about the idea, but about execution at scale.
- Market volatility. An AI can be wrong. A major price forecasting error or an unforeseen grid event could lead to significant losses, eroding customer trust and the promised revenue multiple.
- Customer concentration. The business relies on convincing large, risk-averse battery asset owners and developers to hand over the keys. A single failed deployment with a flagship customer could stall momentum.
- Algorithmic competition. As the space attracts more capital, the differentiation may shift from having an AI to having the best AI. Maintaining a performance edge requires constant R&D investment and access to proprietary data streams.
Rewbi's answer to these risks likely lies in Marge's prior experience. He has seen how storage software companies scale and where they stumble. The focus on ERCOT first is a classic beachhead strategy, dominating one complex, lucrative market before expanding.
The math of a megawatt-hour
The unit economics of battery arbitrage are refreshingly tangible. Take a hypothetical 10-megawatt battery. If Rewbi can consistently capture an average spread of, say, $30 per megawatt-hour between charge and discharge prices, and cycle the battery once per day, that's $300 per day in gross revenue per asset, or about $9,000 per month. The company's claim of doubling a fixed fee suggests they believe their AI can find spreads significantly better than what a naive or manual strategy could achieve. The real test will be whether that margin holds as they aggregate hundreds of megawatts, where their own trading activity begins to influence the very market prices they're trying to exploit.
For now, Rewbi is a small bet with a clear path. It must prove its AI can consistently beat not just human traders, but the established algorithms of incumbents like Stem, the very company that acquired its founder's last venture. The Texas grid, with its wild price swings and open market, is the perfect proving ground to find out if this AI has the spark it needs.
Sources
- [Crunchbase, September 2024] Pre Seed Round - Rewbi | https://www.crunchbase.com/funding_round/rewbi-pre-seed--18b9d2a7
- [getlatka.com, 2026] Revenue reached $330K in 2025
- [LinkedIn, Unknown] Rewbi Company Profile | https://www.linkedin.com/company/rewbi
- [No Cap Blog, Unknown] Thomas Marge - No Cap Blog | https://nocap.blog/founder/thomas-marge/
- [Orange Collective, October 2024] Rewbi Profile | https://orangecollective.co/rewbi
- [Rewbi, May 2025] Rewbi Closes $4m Seed Round | https://www.rewbi.com/post/rewbi-closes-4m-seed-round
- [Y Combinator, August 2024] Rewbi: A stealth mode startup | https://www.ycombinator.com/companies/rewbi
- [Y Combinator, Unknown] Rewbi YC Profile