Top-Three BESS Below 1x Sales

🌾 Welcome to StableBread's Newsletter!

AI data centers are being built faster than the grid can power them. The IEA estimates around 20% of planned data center projects risk delays because grid connections and equipment can't keep up, and expects data center electricity use to more than double by 2030.

GPU clusters (the racks of chips that train and run AI models) spike and drop their power draw within seconds, which the utility grids these data centers plug into were never designed to handle (read my PowerBank write-up). The solution the industry has settled on is placing battery energy storage systems (BESS) between the servers and the grid.

NVIDIA's engineering team now calls BESS "essential infrastructure" for AI factories. NVIDIA also published a formal qualification framework in May for battery vendors on its DSX AI factory platform, which requires proof a vendor can scale manufacturing 10x within 24 months. A requirement like that only makes sense if NVIDIA expects battery orders in the gigawatts.

The market has already repriced the main players selling power into this buildout. Bloom Energy (NYSE: BE), the fuel cell maker providing on-site generation for data centers (like for Nebius (NBIS)), is up 760% over the past year to a $61B market cap. That’s ~16x the $3.7B of revenue analysts expect from Bloom in calendar 2026 ($61B / $3.75B).

Fluence (FLNC) sells the batteries for the same buildout and is guiding to $3.4B of revenue for FY2026 (ends this September), ~$350M under Bloom's calendar 2026 number. Yet Fluence trades under 1x that guidance, with a record $5.6B backlog and order intake running at double last year's pace.

Some of this valuation gap is deserved. Fluence integrates battery cells it buys from others (like AESC and CATL) rather than manufacturing proprietary technology, which is why its gross margins run near 12% while Bloom's run near 30% (more on this later):

The rest of the gap is harder to defend. Fluence is a top-three global battery storage integrator (alongside Sungrow and Tesla) that has signed MSAs with two hyperscalers, and on June 1st Siemens announced an NVIDIA AI factory reference design with Fluence's batteries inside. That means anyone who builds to NVIDIA's blueprint is steered toward Fluence's hardware.

FLNC stock jumped 43.8% on June 1st, the day of the Siemens announcement, touched $30/share on June 2nd, and has since given back the entire move.

The stock now trades around $14.50/share, a $2.67B market cap (184.3M shares × $14.50). That's 48.0% below its June 2nd close.

So is this a mispriced company with huge upside, or just an inferior low-margin integrator the market has priced correctly? That’s what this 6k-word deep dive answers.

Fluence builds grid-scale battery energy storage systems (BESS), the shipping-container-sized banks of lithium-ion batteries that utilities and power developers plug into the grid. The batteries store cheap electricity (usually midday solar) and discharge it when the grid needs it most.

Fluence doesn't manufacture battery cells. It buys them from suppliers at market prices, then wraps its own enclosure design, power electronics integration, controls software, and service contracts that can run 20 years around them.

Management's line on the May call was that Fluence can "make any battery great." The point is that customers aren't paying for the cells. They're paying for the engineering, controls software, and service contracts wrapped around the cells, which is where Fluence's margin lives.

The company was formed in January 2018 as a joint venture between Siemens and The AES Corporation (NYSE: AES), the American power company whose storage arm had been deploying grid batteries since 2007. Qatar's sovereign wealth fund invested $125M in 2020, and Fluence went public at $28/share in October 2021, near the peak of the clean energy bubble.

There are three main revenue lines (% of H1 FY2026 revenue of $940.1M):

The products business is low-margin and capital-hungry. The services and digital contracts are the sticky, recurring side of the business, and management guides ARR to ~$180M by the end of FY2026, up from $148M in FY2025.

Smartstack is the product the entire AI data center story runs through. The platform fits more than 500 MWh per acre, which Fluence calls the industry's highest site-level density, ships initially at 7.5 MWh per pod, and started commercial operation on its first project this spring. In June, Fluence announced a 10 MWh version.

The modular design also accepts pouch cells, the flat-format batteries EV factories are built to produce. That matters for the US supply chain, which I'll get to below.

One structural quirk to understand upfront. Fluence is a "controlled company," meaning its founders hold voting control through a stockholders agreement. AES holds its stake as Class B-1 shares in an Up-C structure, a holding-company setup that gives the founders tax benefits, which Fluence partly pays for under a tax receivable agreement.

What this means is that AES, Siemens, and Qatar decide board seats, not public holders. You get the economics of owning Fluence without the control (unlike normal common stock), so you can't vote out a director and no one can buy the company unless those three agree to sell.

Before we discuss Fluence further, you should understand why the stock fell from $28 at IPO to $4.85 by the end of March 2025.

Short Report and the SEC

In February 2024, short seller Blue Orca accused Fluence of hiding a lawsuit in which Siemens, its own co-founder, alleged engineering failures, false representations, and fraud. The stock crashed, the audit committee of Fluence's board ran an internal investigation with outside counsel and forensic accountants, and Fluence rejected the allegations.

The SEC didn't fully agree that it was nothing. Per the Q2 FY2026 10-Q, the SEC is still conducting a formal investigation into Fluence's financial reporting, though management states it doesn't expect the cost to be material.

Investors sued as well. Two suits filed in early 2025 were consolidated into In re Fluence Energy, Inc. Securities Litigation in the Eastern District of Virginia, covering everyone who bought between the October 2021 IPO and the February 2025 guidance cut, and naming AES alongside Fluence and several executives.

On March 31, 2026, the court dismissed that complaint in full, though without prejudice, meaning plaintiffs were given a window to file an amended version.

So far the SEC investigation hasn't produced anything. Fluence disclosed it in August 2024, the audit committee finished its internal investigation, E&Y has kept signing the accounts with no restatement, and the class action was dismissed.

What keeps it live is the subject matter. On the earnings call that followed the disclosure, CFO Ahmed Pasha said the SEC appears to be focused on revenue recognition, which is the accounting behind 94.1% of Fluence's revenue, and neither the outcome nor the timing is in the company's hands.

Two Guidance Cuts and a Miss

Management entered FY2025 guiding revenue of $3.6-4.4B, cut it to $3.1-3.7B in February 2025 on contract delays (the stock lost nearly half its value the next day), cut it again to $2.6-2.8B in May 2025 after pausing US projects during the tariff chaos, and still missed.

FY2025 closed at $2.26B, down 16.1% y/y, after US factory ramp delays pushed deliveries out.

In August, management had reaffirmed the $2.6-2.8B range and flagged only ~$100M slipping into FY2026. They ended up missing by more than three times that!

Two cuts and a miss in one year. That's the trust deficit this management team is still repairing, and it's why the stock spent close to half of 2025 in the single digits, bottoming at $3.46 in April, even as orders rebounded.

The Legacy Legal Overhangs

Two older legal overhangs cleared in December 2025.

The first was the September 2021 overheating incident at a customer's 300 MW facility (Moss Landing), where Fluence supplied and installed the technology.

A leaking liquid cooling hose set it off, and it's a separate event from the January 2025 fire at the same site. It settled for an amount the company calls immaterial, Fluence's insurers and subcontractors paid part of the settlement, and Fluence admitted no fault.

The second was a customer counterclaim demanding disgorgement of ~$230M, essentially a refund of everything Fluence was ever paid on the Diablo storage project, based on a contractor-licensing technicality. It was dismissed by the court, and Fluence is still pursuing the $37M it says it's owed on that project.

Neither one cost Fluence a material amount, so execution is what to judge from here.

After two cuts and a miss, management needs FY2026 (ending September 30) to land where it says it will.

Fluence guided to revenue of $3.2-3.6B with a $3.4B midpoint, ~50% growth over FY2025, plus adjusted EBITDA of $40-60M and ARR of ~$180M. It has reaffirmed that guidance twice, and as of the February report the revenue midpoint was fully covered by signed orders in backlog.

Management uses three different terms for future business. Don’t get them mixed up:

Now the results, starting with orders.

Fluence signed over $750M of orders in Q1, another ~$574M in Q2, and over $600M in the first five weeks of Q3, putting year-to-date intake at $2B through May 6, double the $1B pace of the prior year:

That's already ~70% of the $2.9B Fluence booked in all of FY2025, with five months still remaining. Backlog hit a record $5.6B on March 31.

Two details in the order book matter more than the headline number:

Pipeline grew to $31.5B as of March 31, up 35% in six months, with the US at 61% of the total:

S&P Global Commodity Insights ranks Fluence among the top three battery storage providers worldwide by installed and contracted capacity, and in July it kept Fluence at Tier 1 supplier status in its 2026 cleantech list. So consider Fluence an incumbent scaling with its market rather than a challenger taking share.

The bigger competitive threat is battery cell makers selling storage systems directly, which I’ll discuss in the domestic content moat section below.

As previously discussed, AI data centers have a power problem that BESS happens to solve. GPU training clusters move their power draw up and down by large percentages within seconds, and a grid interconnection sized for that peak sits underused the rest of the time.

A battery parked between the grid and the servers smooths those swings, provides backup, and lets a data center connect faster with a smaller grid footprint.

The power-quality case for BESS isn't Fluence marketing. NVIDIA's technical explainer and qualification framework set the bar themselves, running 12 tests from millisecond telemetry to islanded operation before NVIDIA lists a vendor's system as qualified for its DSX platform (NVIDIA's blueprint for building AI data centers).

Fluence's Position With the Hyperscalers

So how is Fluence positioned among these data center buyers? Here's what the company has today, in increasing order of commitment:

What Fluence does not have, as of today, is a single data center order converted into backlog or revenue. Management said exactly this on the February call, and in May guided that the first hyperscaler purchase order should land within Q3 of FY2026, which ended June 30. No order has been announced since.

Either it was signed quietly and shows up when Fluence reports Q3 results in August, or it slipped. Both cases leave the AI data center narrative with no revenue behind it today.

None of it sits in the $5.6B backlog either, since backlog only counts signed contracts. Every contract in there today came from utilities and power developers, so a first data center order would land on top of everything above.

Regardless, at ~0.8x this year's revenue, you're paying almost nothing for the data center opportunity.

Goldman Sachs Sizes the Opportunity

Goldman sized the opportunity in a July 16 report. It estimates ~50 GWh of new US behind-the-meter storage demand in 2030, and its utilities team raised its forecast for annual US power demand growth to 3.2% from 2.6% on higher data center demand.

That ~50 GWh in a single year is more than 2.5x everything Fluence has deployed in its history, and 3.2% seems low until you remember US power demand barely grew for the 15 years before 2021.

Goldman puts the behind-the-meter piece alone at 20 GW of new power capacity by 2030, mostly onsite gas. That's a customer base for Fluence rather than competition, since BESS is what sits in front of those generators and absorbs the load swings gas can't follow.

The report's case for BESS comes down to speed. A system deploys in 12-18 months against 4-8 years for grid expansion, making storage the fastest way for a data center to get around the interconnection queue.

The report also names Fluence directly. Goldman held the stock at Buy with a $22 target since May, and the report calls it a meaningful beneficiary of the growing focus on behind-the-meter storage.

If Fluence converts 15-25% of the 12 GWh pipeline at ~$200/kWh, that works out to $360-600M of revenue (1.8-3.0 GWh × $200M per GWh), or 8-14% upside to Goldman's FY2028 revenue forecast.

At the ~0.8x revenue multiple the market pays for Fluence today, that revenue is worth another $1.50-2.60/share (($360-600M × 0.8x) / 184.3M shares), or 11-18% of the current price.

Why Fluence Can Win the Data Center Orders

The hyperscalers' primary requirement, per management, is quality of power, keeping voltage stable while GPU load jumps up and down.

On the February call, CEO Julian Nebreda said these customers require response times below 10 ms for quality-of-power work, against the ~100 ms European transmission codes typically demand.

By May, Fluence said it had developed a proprietary controls solution for these fluctuations, leaning on years of running fast-response systems on European grids.

In June, DNV, the Norwegian assurance firm, confirmed 99.3% availability across Fluence's reviewed fleets of 50 MW and above. But Fluence commissioned DNV and supplied the fleet data, so treat the headline as a company-sponsored review, though DNV states it independently evaluated the assumptions, methodology, and results behind the number.

None of this guarantees the first order. But it answers the question these buyers actually ask, which is whether a vendor can hold power steady at millisecond speed across a real fleet.

Long-Duration Energy Storage

In addition to the data center optionality, long-duration energy storage (LDES) has gone largely overlooked by the market.

LDES means systems discharging six hours or more, and Fluence counted ~34 GWh of these projects in pipeline or leads as of February, mostly in Europe and the US.

Smartstack's density advantage compounds in long-duration configurations because more hours means more containers on the same land.

That leaves Fluence holding two free options on the same product, and the market is really only debating one of them.

Fluence's most defensible asset in the US isn't the product, it's the supply chain.