Eos Energy: the share count is 542 million, not 364

Every stock screener prices Eos Energy at about $1.0 billion: 364 million shares at $2.765. The second quarter 10-Q lists another 178 million shares that cost their owners nothing, or one cent each, to collect: a preferred stock that converts one for one, a warrant with a $0.01 strike and employee stock units. Count them and the equity is worth $1.5 billion, the enterprise about $2.0 billion, for a company that spent $1.71 to build every $1.00 of batteries it sold last quarter.

At the October 8, 2026 close of $2.765 the shares trade 822% above my triangulated Base value of $0.30, a downside of 89.1%. The Bear value is $0, because in that case the business is worth less than the debt and the cash it still has to burn. The Bull value of $2.66 is 3.6% below the price, so the market is paying slightly more than my optimistic case. The required margin of safety (MOS) is 75%, above the framework's 60% ceiling, so no buy price is issued. This is not a verdict on zinc batteries. It's a verdict on what is left for common shareholders after the debt, the preferred stock and the warrants.

Route: SF-08 Pre-FCF growth (enterprise value to revenue), with SF-13 Industrial and Manufacturing for the secondary method. Subtype: zinc-based long-duration battery storage manufacturer. Economic classification: Standard (3/12). Primary method: enterprise value (EV) to 2027 revenue, less 12 months of cash burn. Secondary method: a path-to-cash-flow model on 2029 EBITDA, discounted to today. Values are present fair value estimates, not 12 month price targets.

What Eos is

Eos designs and builds zinc-based batteries for grid storage that runs 4 to 16 hours, a job where lithium-ion is expensive and flammable. The product is the Z3 cube, made in Pennsylvania: Line 1 at Turtle Creek reached an annualized rate of about 1.5 GWh in June, and Line 2 at Thorn Hill started commercial production in mid-June on one partial shift. Eos is moving all production to Thorn Hill, which the Department of Energy (DOE) loan supports at 4 GWh a year. Backlog was $807 million on June 30, or 3.4 GWh, which works out to about $237 per kWh.

The customer list changed this summer. Eos formed Frontier Power USA (FPUSA), a project developer, with funds managed by Cerberus and Hudson Bay. Eos put in $112.6 million for about 36% of it, raised through a $75 million registered direct offering and a rights offering. FPUSA then buys Eos batteries for its projects. A pre-existing project financed by a Cerberus affiliate produced $55.0 million of second quarter revenue, 80% of the total, and on June 30 that project and FPUSA made up 49% of backlog volume. Cerberus is also Eos's largest lender and holds the preferred stock and the penny warrant.

The latest quarter

The low end of guidance needs the second half to deliver about $174 million, against $126 million in the first half. Analysts sit at $304 million for the year, below the $325 million midpoint (third-party consensus). The range was narrowed because Line 1 goes offline while it moves to Thorn Hill.

Why the gross margin is worse than -71%

The -71% already includes a subsidy. Eos earns a refundable production tax credit of $35 per kWh for battery cells and $10 per kWh for modules, and books it as a reduction of cost of goods sold: $12.5 million in the second quarter, 18.1% of revenue. Before the credit, the gross margin was -89.1%. Under current law the credit runs at full rate through 2029 and then phases down through 2032, so the margin that matters for a long-term value has to work without it.

Net income is not a usable metric either. Eos reported a net profit of $233.2 million for the first half of 2026 while losing $163.1 million from operations. The difference is mostly fair value gains on warrants and derivatives, which shrink when the share price falls. The worse the stock did, the better the bottom line looked. I use revenue, gross margin and free cash flow (operating cash flow minus capital spending), which was -$262.2 million in the first half, about -$131 million a quarter.

Management is targeting more than 72 points of adjusted gross margin improvement over the next 12 months, which would take adjusted gross margin from about -62% to about +10% by mid-2027, a level an analyst cited on the call. Adjusted gross margin was -62.3% in the second quarter and excludes $5.9 million of depreciation and stock compensation in cost of goods sold.

Debt, preferred stock and the real share count

The balance sheet shows $617.1 million of borrowings. The amount owed is $948.3 million: $600 million of 1.75% convertible notes due 2031 (convertible at about $16.16, carried at $317.9 million after a $281.8 million discount), $50 million of 6.75% convertible notes due 2030 (convertible at about $5.06), $96.9 million of DOE loan including capitalized interest, and $201.4 million of a related-party term loan from a Cerberus affiliate, carried at fair value of $163.3 million. In September Eos drew another $87 million from the DOE loan, bringing total DOE draws to $178 million. Pro forma for that draw, debt is $1,035.3 million and unrestricted cash $392.5 million, so net debt is $642.8 million before third quarter cash use. Restricted cash of $58.6 million is left out, because part of it is reserved for interest on the notes.

The share count needs the same treatment. The Series B preferred stock participates with common shareholders as if converted and was carried at $713.2 million on June 30, because its redemption value is the converted shares times the share price. It is equity, not a senior claim. The SPA warrant held by a Cerberus affiliate is a single warrant for 44.4 million shares at $0.01. Both were adjusted upward by the summer offerings.

That is 48.9% more claims than the 364 million on the screens. Three items are left out: 15.96 million stock options, whose exercise prices I could not verify, Hudson Bay's right to exchange its FPUSA units for up to 9.12 million shares at $5.481 or more, and the convertible notes, which are far out of the money and sit in debt. A registration of 56.5 million shares in August covers resales of shares from the preferred stock, the penny warrant and the $5.481 warrants listed above. It is not new dilution.

Primary valuation: enterprise value to 2027 revenue

With negative gross margin there are no earnings or cash flows to value, so the framework for pre-profit companies uses EV to forward revenue. Its speculative tier is 1x to 3x. Two listed storage peers frame it: Fluence trades near 0.55x trailing revenue with a 9.4% gross margin, and Energy Vault near 3.8x trailing with a 22.5% gross margin (StockAnalysis, October 2026). I use 2.0x for Eos: a premium to Fluence for growth of 70% or more and a domestic, non-lithium product, and a discount to Energy Vault for a gross margin that is still deeply negative. It is an assumption, not a derived figure, and the grid below tests it.

For 2027 revenue I use $540 million, 6% below the $575.6 million consensus (my estimate), because the company missed its 2025 range by a wide margin and the 2026 consensus already sits below guidance. That gives an EV of $1,080 million. Then: less net debt of $642.8 million, less $300 million of cash burn over the next 12 months (my estimate, against $262.2 million in the first half alone), plus the FPUSA stake at its $112.6 million cost. Equity is $249.8 million, or $0.46 a share on 542.1 million claims.

Sensitivity: value per share

All cells use $300 million of 12 month burn and the FPUSA stake at cost. Equity is a thin slice on top of $943 million of debt and burn, so every 0.5x of multiple moves the value by about $0.50 a share. Only 4.0x on revenue at or above consensus clears the price.

Secondary valuation: path to cash flow

The secondary method asks what Eos is worth once it is a normal manufacturer. In the Base case, 2029 revenue is $950 million, about 4.0 GWh at $237 per kWh, which is the full Thorn Hill capacity. EBITDA margin is 15%, which still includes the production tax credit that phases out after 2029. At 12x EBITDA, the top of the standard industrial range of 8x to 12x, EV at the end of 2029 is $1,710 million, or $1,086 million today at a 15% discount rate over 3.25 years. Until then Eos has to fund the losses: about $200 million in the second half of 2026, $200 million in 2027 and $80 million in 2028, so $480 million in total (my estimates). Less net debt and burn, plus the FPUSA stake, equity is $75.8 million, or $0.14 a share.

Sensitivity: value per share at $950 million of 2029 revenue

At a 12% discount rate instead of 15%, the Base value is $0.32.

Market multiples and growth diagnostics

A historical valuation cross-check is not meaningful: revenue was $15.6 million in 2024 and $114.2 million in 2025, so any multiple history describes a startup, not the company being valued. At $2.765 and 546.6 million share claims, EV is $2,042 million, net of the FPUSA stake. That is 6.3x the 2026 guidance midpoint and 3.5x 2027 consensus revenue. The screens show trailing EV to revenue of 6.1x on an EV of $1.30 billion, because they use the carrying value of debt and 364 million shares.

The growth valuation cross-check (SM-19) does not apply: earnings are negative and distorted by fair value swings, so P/E and PEG are not meaningful (PEG_NOT_MEANINGFUL). Revenue growth is structural with high execution risk; cyclical distortion is low, because the risk here is manufacturing and funding, not a demand cycle. The free cash flow growth check (SM-20) starts at -$262.2 million for the first half of 2026. A growth rate from a negative base is not meaningful, free cash flow per share is negative, and the low base flag is high. In my Base path free cash flow turns positive in 2029, so the forward free cash flow yield is negative for the next two years. Estimate reliability: low. Classification: uncertain, path-to-cash-flow. SM-19 and SM-20 do not set fair value, MOS or the verdict.

Valuation triangulation

At the level of the whole company the two methods agree almost exactly: $1,080 million and $1,086 million of EV, a 0.5% difference. Per share they differ by 107%, because the primary method deducts 12 months of burn and the secondary deducts 30 months, and equity is a small residual where $180 million is a third of a dollar. Formally that is a valuation divergence flag, and it lowers confidence and raises the margin of safety. Both methods carry estimated inputs, so I weight them equally. The Base is $0.30.

What is the market already pricing in?

The price needs 4.3x my 2027 revenue after 12 months of burn, or 2027 revenue of $1,171 million at my 2.0x. In the secondary model it needs 2029 EBITDA of $331 million at 12x: a 34.8% margin on $950 million of revenue, or 25.5% on $1.3 billion, which would mean expanding beyond the 4 GWh plant. For a product that has never earned a positive gross margin and whose margin includes a tax credit worth 18% of revenue, that is the best outcome on the table, not a central one. Short interest is 33.6% of the float (StockAnalysis), so the price can still move sharply in either direction.

Bear, base and bull scenarios

The Bear values the FPUSA stake at half its cost; in both methods equity comes out negative and is floored at $0. The Bull adds in-the-money warrants by the treasury method (544.8 million and 547.7 million claims). Revenue in all three cases sits inside or near the range that guidance and consensus allow for 2027; the cases differ mostly in margin and funding. I assign no probabilities.

How the buy price is set

A Confidence D score carries a base MOS of 50%. I add 10 points for leverage and funding needs, 5 for customer concentration, 5 for accounting opacity (fair value swings, a related-party lender who is also a customer's sponsor) and 5 for the per-share divergence between the methods. The required MOS is 75%. Above 60% the framework prefers no buy price over a pseudo-precise one, so no buy zone is issued. For reference, 75% off the $0.30 Base would be about $0.07.

Dividend and capital return

Eos pays no dividend and buys back no stock. Capital flows the other way: basic shares rose 42.9% in a year, and the 2026 offerings added warrants for 39.1 million more shares at $5.481.

Reasons to own EOSE

• A real demand signal. Backlog of $807 million is 2.5 times the 2026 guidance midpoint, and a Department of War contract for the Golden Dome program adds a defense customer.
• Non-lithium, domestic supply. Zinc chemistry avoids thermal runaway and Chinese cell supply, which matters for US content rules and the production tax credit.
• Visible cost levers. Line 2 runs about 10% faster cycle times than Line 1, material cost per cube fell 10% in the quarter, and consolidation into Thorn Hill is expected to cut conversion costs by 10% to 15%.
• Long-dated, cheap debt. The $600 million notes carry 1.75% to 2031, the DOE loan runs to 2034, and the related-party loan can capitalize interest.
• Optionality in FPUSA. A 36% stake in a developer with about 16 GWh of pipeline could be worth more than its $112.6 million cost if projects get built.

What could go wrong?

• Thesis breaker: gross margin stays negative. The -71% already includes a tax credit worth 18% of revenue. If the cost curve stalls, every extra GWh adds to the loss.
• Funding and dilution. Pro forma unrestricted cash of $392.5 million covers about 3 quarters at the first half burn rate. The next raise, at today's price, dilutes a share count that already has 542 million claims.
• Concentration and related parties. 80% of second quarter revenue came from one Cerberus-financed project, and FPUSA, partly funded by Eos shareholders, is now a key buyer. Cerberus is also lender, preferred holder and holder of the penny warrant.
• Covenants and maturities. The DOE and related-party loans carry minimum liquidity covenants, and DOE revenue and EBITDA covenants start on March 31, 2027. A springing maturity could pull the DOE debt forward to March 2030.
• Policy. The production tax credit phases down after 2029, and customers' project economics depend on federal incentives.
• Execution. Line 1 goes offline during the move to Thorn Hill, and the top of 2026 guidance needs Thorn Hill running around the clock by late fourth quarter.

Management execution

In August 2025 management reaffirmed 2025 revenue of $150 million to $190 million. The year came in at $114.2 million, 24% below the low end. The 2026 range was narrowed from $300 million to $400 million to $300 million to $350 million, and analysts are below the midpoint. On the other side, Line 2 started on schedule, revenue in the first half of 2026 exceeded all of 2025, and the cost initiatives are measurable. Capital allocation has relied on related-party financing and a 42.9% increase in basic shares in a year. Assessment: 6/20, weak. The next test is the third quarter report, expected on November 4, 2026.

Stock Analyza scorecard

Revenue growth scores 2/2 mechanically, from a very low base. Everything that depends on profit or cash scores zero. Business quality is 23/100: moat and pricing power 7/20 (proprietary chemistry and domestic supply, but no proof of a cost advantage), return versus cost of capital 0/20, balance sheet 2/15, earnings and cash quality 1/15, growth and runway 9/15, management and capital allocation 3/10, dilution and governance 1/5. The valuation score is 7/100: discount to Base 0/40, protection against the Bear case 0/20, agreement between methods 5/15 (close at the EV level, far apart per share), historical valuation 0/10 (not available), reverse valuation 0/10 (price above Bull), data and model quality 2/5.

Confidence score

The capital structure is fully disclosed in the 10-Q, which is why data quality scores 4. Everything else is held down by a business that has not shown positive unit economics, a value that swings $0.50 a share for each 0.5x of multiple, and a range of $0 to $2.66 between Bear and Bull.

What I would watch from here

Today four readings are red: burn (about $131 million a quarter in the first half), cash coverage (about 3 quarters, pro forma for the DOE draw), related-party revenue (80%) and share growth (basic shares +42.9% in a year). Gross margin is yellow: -71% in the second quarter against about -78% in the first, an improvement of 7 points. Revenue against guidance is yellow, with consensus at $304 million. Production is yellow, since Line 2 started on time but runs one partial shift. The next review point is the third quarter report, expected on November 4, 2026.

The three most important thesis breakers

1. Gross margin stays negative through 2027 despite the Thorn Hill ramp.
2. A financing round at a low price adds materially to the 542 million share claims.
3. FPUSA or the related-party projects fail to convert backlog into revenue.

Adversarial review

I recomputed every value in this article in a separate model and checked the inputs against the second quarter release, the 10-Q notes on borrowings, warrants, preferred stock, earnings per share and subsequent events, the August resale registration and the closing price. The capital structure inputs are reported. The multiples, revenue paths, margins and burn are my estimates, and the per-share result depends on them more than in a typical analysis, because in my Base equity is only 7% to 23% of a company valued at about $1.1 billion.

The weakest points of the model: the 2.0x multiple is an assumption bracketed by two peers rather than a peer table; the burn path to 2029 is an estimate; the FPUSA stake is carried at cost with no market price; and 15.96 million options and the Hudson Bay exchange right are excluded. The direction of every item I could not quantify is toward more dilution, not less.

The strongest counter-thesis, for the bulls: Eos owns the only scaled US zinc battery line, demand for long-duration storage is growing with data centers and defense, and if Thorn Hill pushes gross margin to 30% or more while capacity grows past 4 GWh, the company could earn $300 million or more of EBITDA by 2029. The response is that this is the Bull case, it gives $2.66, and today's price is above it. The most generous secondary cell I tested, 25% margin and 16x on $1.3 billion of revenue, gives $4.45.

Audit flags: 2026 and 2027 revenue consensus come from StockAnalysis and were used only as a reference point, not as the Base. The peer multiples are trailing figures from the same source. The adjusted gross margin target is management's. The third quarter cash use is not yet reported, so the net debt figure is as of June 30 plus the September DOE draw. Arithmetic: Q2 gross margin -48.801 / 68.775 = -70.96%; related-party share 55.034 / 68.775 = 80.02%; basic share growth 339.799 / 237.741 - 1 = 42.93%; share claims 542.09 / 364.17 - 1 = 48.86%. The publication gate status is PASS WITH WARNING: the capital structure is reconciled, and the value rests on estimated margins, multiples and burn.

Final verdict: sell at $2.765

Eos has a differentiated product, a large backlog and a factory that is getting faster. It also has $948 million of debt, a tax credit hiding 18 points of gross margin, its biggest customer funded by its biggest lender, and 542 million share claims where the screens show 364. At my Base the business is worth about $1.1 billion and almost all of it belongs to the lenders and the burn. The stock price is only cheap if you count a third of the owners out.

Verdict: SELL at $2.765. Triangulated Base value $0.30, downside 89.1%, Confidence D (9/25). Required margin of safety 75%, no buy zone.

The case changes if GAAP gross margin turns positive with the production tax credit no larger than today, if new orders come from customers outside FPUSA and Cerberus, or if the next financing is done without adding to the share count.

Sources

Primary sources include Eos's second quarter 2026 release of August 5, 2026, the Form 10-Q for the quarter ended June 30, 2026 (notes on borrowings, warrants, government grants, redeemable preferred stock, earnings per share and subsequent events), the resale registration of August 31, 2026, the registered direct offering prospectus of June 30, 2026, the DOE loan advance release of September 14, 2026, the second quarter 2025 release with the 2025 guidance, and the second quarter 2026 earnings call. The closing price, consensus estimates, short interest and peer multiples come from StockAnalysis.

The reference market price is the October 8, 2026 regular session close of $2.765. Balance sheet figures are as of June 30, 2026, adjusted for the $87 million DOE draw in September; the common share count of 364.17 million is as of August 3, 2026. The multiples, revenue and margin paths, cash burn and discount rate are analytical estimates, not company guidance.