Invinity Bets Its Comeback on AI Data Centers — Just as the Tape Sours on Them
The reframing: from LDES niche to 'flexible' AI infrastructure
Invinity Energy Systems plc (IES.L, ~£128m market cap) reported interims on 2026-09-22, and the message from CEO Jonathan Marren was deliberately rebuilt from the ground up. Rather than selling a vanadium Flow Battery as a long-duration-energy-storage or "non-lithium" play, he now pitches it as a flexible asset: “what we offer is a flexible energy storage solution” — Jonathan Marren, Chief Executive Officer · 2026-09-22. The argument is that a 30-year, cycle-unlimited battery never becomes a stranded asset in a market where grid requirements are being rewritten by electrification and, above all, by data center loads.
That pivot is quantifiably new. In the prior year's Q&A, the data center case was still exploratory — Matt Harper describing loads “all over the place” — Matthew Harper, Senior Executive or Strategy/Commercial Lead · 2025-10-10 as interesting rather than addressable. This quarter it is a numbered market: a AI data centers opportunity framed at ~£45bn, with a claimed 15% cost advantage on inference sites and 12% on training sites. The headline proof point is the FlexBase project in Switzerland — 1.5 GWh, sized by Harper as roughly 1.5x the average load of Zurich, with installation beginning "in about a year or so's time" for 2029/2030 delivery.
The cost curve is the real unlock
What makes the AI pitch credible is not the idea but the product cost work. Management cites ~60% cost-out on the legacy VS3 platform and ~66% on the new Endurium design for early-2027 shipments, with a further ~50% targeted by 2030. The specific quote that crystallizes the commercial shift is an EPC partner's feedback that Invinity was the Rolls Royce of vanadium flow batteries — meaning premium quality and premium price. The bet is that as price points fall, previously closed doors open. Adam Howard's numbers show early traction: “Customer orders that we've signed there are up 3x year-on-year, to just over 30 MWh” — Adam Howard, Chief Financial Officer · 2026-09-22, with an order book just under 80 MWh and GBP 10m cash against GBP 40m of inventory built for H2 delivery.
A company-unique wound: the Cap-and-Floor snub
The most interesting — and most Invinity-specific — element is the Cap-and-Floor disappointment. The UK's flagship long-duration scheme awarded consent to pumped hydro and lithium projects rather than flow batteries, and the share price, per Marren, had been performing well right up to that announcement. He is unusually blunt about fighting the outcome:
My view is I would prefer. Our share price was doing very well up until that announcement... Will that ultimately result in a change. I do not know and I cannot forecast that. But I tell you what, I would far prefer to fly the flag of Invinity and do our best to get some change than hide away.
This matters because Cap-and-Floor was the single biggest credibility anchor in prior calls — a scheme repeatedly sized at 16.7 GWh, billions of pounds, and the reason for building UK stack manufacturing. The scheme's rejection pushes Invinity toward the US, where it is "most of the way through selecting a site" to capture the ~50% stacked ITC/PTC incentive package. Note the contrast with prior calls, where management stressed scale-up in the UK Central Belt; the emphasis has rotated hard toward a US factory.
Confluence check: riding a wave the market has already sold
Globally, AI data centers is a megatheme — but the tape has been fading it. The theme sits +66% on a 360-day window yet is down double digits over 90 days and further over 30 days, with sub-themes like HPC data centers, data center power and data center capacity all in the decliner lists. Invinity is pivoting into a narrative the market has, for now, been de-rating. That is a genuine disagreement worth flagging: the company is late-cycle-early on a cooling theme.
Where the company is more fortunate is in the California Energy Commission channel — a state-level backer that keeps steering projects (Pacific Steel, Indian Energy) its way even as federal policy churns. It is a company-unique, durable demand signal that does not depend on the AI hype cycle.
The read
What changed: a cost curve crossed a threshold, a 1.5 GWh anchor project was won, and a genuinely new segment (AI facility buffering) was quantified. What did not change: a small-cap with GBP 10m cash, a going-concern note leaning on inventory conversion, and a UK policy path that just got harder. The honest framing is that Invinity has a stronger product story and a weaker demand-policy story than it did a year ago. Management is asking investors to underwrite a curve — cost down, price point down, pipeline converting — rather than a single order. Given 2023's pathway to profitability language and 2025's pipeline-rebuild promises, that is a fair thing to be sceptical of. But the Copwood commissioning (Q4) and the US site announcement are concrete near-term markers to test the story against.