Why Goldman Sees a British Fuel Cell Maker as Its Biggest Power Stock Upside Call
A Bet That Doesn’t Fit the Usual Pattern
Wall Street analysts do not, as a rule, put triple-digit return forecasts on companies that lose money and are expected to keep losing it. That kind of number belongs to speculative biotech, not industrial licensing businesses. So when Goldman Sachs named a small British fuel cell developer as one of its highest-conviction picks in Europe and attached a 168% expected gain to it, the call stood out precisely because it broke that pattern. This power stock upside forecast is one of the largest Goldman has issued on any name in Europe this year.
The company is Ceres Power, a U.K. firm most casual investors have never heard of, and the number attached to it demands an explanation that goes beyond a headline.
The Month That Set the Price
Ceres shares fell more than 31% in July, part of a broader pullback as investors grew nervous about the pace and payoff of artificial intelligence spending. Chipmakers, data center operators, and the smaller companies feeding into that supply chain all took a hit as the market reassessed how quickly AI investment would translate into revenue.
Goldman looked at that same drop and reached the opposite conclusion. Rather than reading it as a warning sign, the bank’s analysts treated the sell-off as a discount on a stock they already liked. That is the entire mechanical origin of the 168% figure: Ceres closed around 383p on August 5, and Goldman’s price target of 930p, set by analyst Michele Della Vigna, sits far enough above that closing price to produce the number making headlines. The stock is up more than 168% year to date even after the drop, yet it remains well below its 52-week high of 872.50p, which tells you how sharply sentiment swung in a single month.
What Ceres Sells
Ceres does not generate electricity, and it does not build power plants. It designs solid oxide fuel cell technology, a system that converts fuel such as natural gas or hydrogen into electricity through a chemical reaction rather than combustion, and then licenses that design to manufacturers who build the physical systems.
That distinction matters more than it sounds like it should. A licensing model means Ceres never has to build factories, carry heavy manufacturing overhead, or manage physical supply chains at scale. Analysts describe this as an asset-light approach, and Goldman expects it to push gross margins above 80% as royalty revenue scales. That kind of margin ceiling is rare in industrial hardware, and it is the financial engine behind the entire bull case.
The Problem AI Created for Itself
Data centers running AI models consume enormous amounts of electricity, far more than a typical cloud computing facility, and the physical grid has not kept pace. Utilities cannot build new transmission lines and substations fast enough to match the speed at which AI companies want to add computing capacity. Goldman expects global data center power demand to rise by roughly 160% by 2030, a figure that reframes electricity access as a bottleneck on AI growth itself, not just a cost line.
Solid oxide fuel cells offer a way around the wait. Because they generate power on-site through a chemical process rather than pulling from the grid, a data center operator can add capacity without waiting years for a new grid connection to be approved and built. That speed is the entire reason a small U.K. licensing company is suddenly relevant to the biggest infrastructure story in technology.
Money That Depends on Someone Else’s Factory
Here is the part of the story that gets skipped when a 168% number leads the headline: Ceres does not manufacture anything, which means its revenue is entirely dependent on what its licensing partners choose to do. South Korea’s Doosan Fuel Cell and Taiwan’s Delta Electronics are both developing and mass-producing fuel cell systems built on Ceres technology, with data centers as the explicit target market.
That arrangement creates a very specific kind of risk that has nothing to do with AI demand or fuel cell technology itself. If Doosan or Delta slow down their production plans for any reason, budget constraints, supply chain trouble, a shift in corporate priorities, Ceres royalty income slows with them. It does not matter how strong AI power demand looks in an industry forecast if the two companies actually building the hardware fall behind schedule.
Ceres and Delta have targeted the start of commercial production by the end of 2026. That date is not a footnote. It is the single clearest checkpoint investors have for judging whether the entire thesis is tracking toward Goldman’s number or drifting away from it.
The Years Before the Payoff
Goldman’s own numbers tell an uncomfortable second half of the story. The bank’s analysts have predicted negative earnings per share for Ceres through 2027, even while forecasting strong long-term growth. That is not a contradiction so much as a description of the bet itself: investors buying Ceres today are paying for royalty income that mostly arrives between 2028 and 2030, not for anything the company is earning right now.
Three things have to go right, in sequence, for the 168% target to actually materialize. Doosan and Delta need to hit their data-center production timelines. Royalty revenue needs to scale fast enough to lift margins toward the 80%-plus range Goldman is modeling. And AI-driven power demand needs to stay strong for the rest of the decade, not just for the next earnings cycle. Slip any one of those, and the target date moves later along with the return.
A List Full of Names, One Outlier
Goldman placed Ceres near the top of its August European Conviction List, the bank’s roster of its highest-confidence Buy calls, and of every name on that list, Ceres carries the largest forecasted gain. That alone tells you something about how the bank is weighing risk against reward across its coverage universe. A 168% forecast is not a rounding error or a modeling quirk; it is the outlier the bank chose to lead with.
Della Vigna is not new to this call. He was among the first major analysts to explicitly connect fuel cell technology to AI power demand, and he has raised his Ceres price target repeatedly through 2026. That track record does not guarantee the target is correct, but it does mean the thesis has been tested and adjusted over time rather than assembled overnight.
Bloom Energy Already Ships Power to Data Centers
Ceres is not the only way to invest in AI’s electricity problem, and the comparison to its closest peer is where the extra risk in Goldman’s call becomes visible. Bloom Energy, a U.S. rival, already supplies fuel cells to data centers through partners including Oracle, and Morgan Stanley has noted that Bloom can withstand a delay in an Oracle project without the thesis collapsing.
That distinction is the whole ballgame. Bloom has real, present-tense revenue from operating fuel cell installations. Ceres does not, at least not yet at meaningful scale. Larger power names such as GE Vernova and Constellation Energy offer exposure to the same AI electricity demand story, backed by real, current profits, which makes them steadier investments but far less explosive ones. Ceres sits at the opposite end of that spectrum: the largest forecasted gain on Goldman’s entire European list, paired with the least proven earnings of any name mentioned alongside it.
What Would Confirm the Thesis by Late 2026
A 168% price target from a single analyst, however well-informed, is a forecast, not a guarantee. It reflects a genuine, well-reasoned read on where AI power demand is headed and how a licensing model could benefit from it, but it depends on partners executing on schedules they do not fully control and revenue that has not yet arrived.
For an investor comfortable with real risk, Ceres offers something the larger, steadier power names don’t: direct, front-loaded exposure to the 2026 through 2030 data center buildout, at a price the July sell-off already brought down. Position sizing matters enormously here, because a pre-profit stock with this kind of forecast attached can fall as fast as it climbed. For anyone less comfortable with that kind of swing, the more useful move may simply be watching the calendar. If Doosan and Delta begin commercial production on schedule by late 2026, that is a concrete, verifiable data point the current price does not yet reflect, and it is the moment the thesis either starts proving itself or starts running out of time.
FAQ
Why did Goldman Sachs give Ceres Power a 168% upside target?
The 168% figure comes from comparing Ceres’ closing price of about 383p on August 5 against Goldman analyst Michele Della Vigna’s 930p price target, set on July 4. Goldman views the stock’s July sell-off as a buying opportunity tied to rising AI-driven electricity demand rather than a signal of weakening fundamentals.
How does Ceres Power make money if it doesn’t build fuel cells itself?
Ceres designs solid oxide fuel cell technology and licenses it to manufacturers, earning revenue through licensing fees and royalties rather than factory sales. This asset-light structure is why Goldman expects gross margins to climb above 80% as royalty income scales with partner production.
What is the biggest risk to the Ceres Power investment thesis?
Ceres’ revenue depends entirely on manufacturing partners Doosan Fuel Cell and Delta Electronics hitting their production timelines. If either company slows its rollout of data-center fuel cell systems, Ceres’ royalty income slows regardless of how strong overall AI power demand remains.
How is Ceres Power different from Bloom Energy?
Bloom Energy already generates real revenue supplying fuel cells to data centers through partners like Oracle, while Ceres has not yet reached that scale. Ceres offers a licensing-based model with larger theoretical upside but less proven, current earnings than Bloom.
When will investors know if the Ceres Power thesis is working?
The clearest milestone is the end of 2026, when Ceres and Delta Electronics have targeted the start of commercial production for data-center fuel cell systems. Hitting that timeline would give the thesis a concrete data point that the current stock price does not yet reflect.
Related Reading
- The Real Reason AI Power Demand Is Straining the Grid (It’s Not What You Think)
- Why LS Power Just Bought a Texas Gas Plant Nobody Was Selling Six Months Ago
Official Source Links
For further reading and verification:
