AI’s next constraint: electricity, not GPUs
The last two years were defined by a scramble for GPUs. The next two will be defined by a scramble for electrons. Google’s new investment in Redwood Materials is not really about recycling feel‑good stories; it’s about securing power for an AI future that current grids are not ready to deliver.
In this piece, we’ll unpack why Google and Nvidia are quietly buying into battery recycling and second‑life storage, how this reshapes the economics of data centres, what it means for European energy and climate policy, and why “battery infrastructure” is becoming as strategic as semiconductor fabs.
The news in brief
According to TechCrunch, U.S.-based Redwood Materials has extended its Series E funding round to $425 million. The round, first announced in October and led by Eclipse, has now attracted new strategic money from Google, joining Nvidia’s venture arm NVentures and existing investors such as Capricorn and Goldman Sachs.
The updated round reportedly values Redwood at more than $6 billion post-money and brings its total capital raised to around $4.9 billion. Founded by former Tesla CTO JB Straubel, Redwood started as a battery recycling specialist, processing scrap from electric vehicle (EV) and consumer electronics batteries and supplying recovered materials like nickel and lithium back into the battery supply chain.
More recently, the company launched “Redwood Energy”, a business that repurposes EV battery packs for stationary energy storage. These systems can form microgrids that power AI data centres and industrial sites. TechCrunch notes Redwood already handles over 70% of used battery packs in North America and targets 20 GWh of grid‑scale storage deployed by 2028.
Why this matters
This funding is not just another climate-tech round; it is a signal about where hyperscalers see their next systemic risk. The constraint on AI growth is shifting from scarce GPUs to scarce, reliable, and green electricity.
For Google and Nvidia, backing Redwood is a hedge on that risk. If you expect AI workloads to multiply, you either depend on utilities and slow grid upgrades, or you start to vertically integrate energy, at least at the edges. Second‑life batteries sitting next to data centres give operators three advantages:
- Capacity: They can build additional compute in places where the grid connection would otherwise be a hard cap.
- Cost: They can arbitrage time-of-use pricing and avoid expensive peak power, which matters when training large models.
- Carbon: They can store surplus renewable energy and smooth intermittency, supporting corporate net‑zero claims.
Redwood benefits by moving up the value chain. Recycling is a margin‑thin, regulation‑driven business. Turning used EV packs into turnkey microgrids for blue‑chip customers is closer to infrastructure and software margins than to scrap processing. If it executes well, Redwood stops being a “waste” company and becomes a critical supplier to the AI boom.
The losers, at least in the medium term, may be conventional utilities and grid operators who assumed hyperscalers would always be captive customers. If more data centre capacity is self‑buffered by behind‑the‑meter storage, traditional grid planning and revenue models will look increasingly outdated.
The bigger picture: AI, power and circular hardware
Redwood’s pivot lands at the intersection of three powerful trends.
1. AI is blowing up power forecasts.
Across the U.S. and Europe, utilities have revised data centre demand projections sharply upward. Training frontier models and running generative AI at scale is simply far more energy‑intensive than the cloud and search traffic of the 2010s. Grid upgrades, transmission lines and new generation all take 5–10 years to deliver. Startups that can “manufacture” virtual grid capacity via storage in 2–3 years suddenly look very attractive.
2. Second‑life batteries are finally finding a real market.
For a decade, the industry has talked about repurposing EV packs that are no longer good enough for cars but are still perfectly fine for stationary storage. The bottleneck was always logistics, safety, and bankability. Redwood has three things few others have at scale: access to a large flow of used packs, technical know‑how from recycling, and enough capital to standardise, certify and insure these systems for conservative customers like data-centre operators.
3. Circular supply chains are maturing.
Redwood started on the materials side – closing the loop on lithium, nickel and cobalt. With Redwood Energy, it is also closing the loop on the hardware lifecycle. The same company now touches a battery when it is born (via cathode materials), during its “retirement” (recycling), and during its second career as stationary storage. That kind of integration makes regulators happy and locks in customers who care about traceability and ESG reporting.
Competitively, this move also puts Redwood into more direct overlap with Tesla’s Megapack business, CATL’s grid‑scale batteries, and a long list of energy‑storage developers. The differentiator is not just technology but feedstock: having first call on 70% of discarded battery packs in North America is a structural advantage that pure‑play storage rivals simply do not have.
The European and regional angle
For Europe, Redwood’s strategy hits a nerve: AI ambitions are colliding with fragile power systems and aggressive climate targets.
European data centres already cluster in places like Ireland, the Netherlands, Frankfurt and the Nordics, where cooling and connectivity are favourable. Several of these regions are running into local opposition and grid constraints. Ireland’s regulator has effectively paused new connections around Dublin; the Netherlands has pushed back on hyperscale projects due to land and energy pressure.
EU policy is simultaneously pushing more electrification (EVs, heat pumps, green hydrogen) and more digitalisation (AI, cloud). The new EU Batteries Regulation demands high recycling rates and detailed material traceability, while the Green Deal and Fit for 55 package require rapid emissions cuts. Second‑life storage sits neatly at this intersection: it helps balance renewable-heavy grids, extends the useful life of imported battery materials, and offers a lower‑carbon way to expand data-centre capacity.
Europe already has strong players in battery materials and recycling – think Northvolt in Sweden, Umicore in Belgium, Verkor in France – and an increasingly vocal AI and cloud ecosystem. Yet there is still a gap where Redwood is now moving aggressively: tying second‑life storage explicitly to digital infrastructure.
For European operators, this raises a strategic question: do they want to depend on U.S. suppliers for both chips and the batteries that keep their data centres running, or should they build a native ecosystem around circular energy storage attached to AI and cloud hubs? Given Europe’s sensitivity to energy security and industrial policy, this is unlikely to remain a niche question.
Looking ahead
Several things are worth watching in the wake of Google’s bet on Redwood.
From pilot to platform. How quickly can Redwood turn its battery inventory into standardised products and software? Hyperscalers do not want bespoke engineering projects; they want replicable, modular solutions they can copy‑paste across campuses.
Commercial structures. Will Redwood own and operate these microgrids, selling power as a service, or mainly sell hardware? The former would turn it into an infrastructure operator with long‑term, utility‑like revenue – but also higher capital needs and regulatory exposure.
Regulation and local pushback. As data centres become “quasi‑power plants” with big batteries on site, expect tighter scrutiny from fire safety authorities, grid regulators, and local communities. Europe in particular will demand rigorous standards for second‑life packs.
Copycats and consolidation. Expect similar moves from other recyclers and from utilities themselves. Oil & gas majors repositioning as “integrated energy companies” will not ignore a chance to own both electrons and the AI workloads that consume them.
In the near term (2–3 years), the more likely outcome is a patchwork: a few high‑profile campuses running on hybrid setups mixing grid power, on‑site renewables and second‑life storage, while most facilities still depend on diesel backup and traditional connections. By the end of the decade, though, any serious AI cluster without significant local storage will look outdated.
The bottom line
Google’s investment turns Redwood from a recycling success story into a front‑line player in the AI infrastructure race. The message is clear: if AI is the new oil, then secure, low‑carbon electricity is the new pipeline – and second‑life batteries are one of the fastest ways to build it. The open question for readers, especially in Europe, is whether you are comfortable letting a handful of U.S. tech firms and their favoured suppliers own both your data and the power systems that keep it alive.
