Anduril is building a prime. The price assumes a platform

Anduril raised $5 billion at a $61 billion valuation

The completed May financing raised $5 billion at a $61 billion valuation; Anduril reported $2.2 billion of 2025 revenue. Reuters subsequently reported discussions about financing near $100 billion, potentially with performance conditions. I use $61 billion as the completed transaction reference and $100 billion to test the expectations in those reported discussions.

March press reporting put projected 2026 sales at $4.3 billion and operating losses at $1.2 billion. Those are March planning figures. The sales anchor is seven months old; new awards do not automatically raise recognized revenue. The secondary revenue report repeats The Information without independent confirmation.

The unknowns include product contribution, funded orders, customer financing, acceptance requirements, warranties and shareholder rights. They are connected: an advance can reduce cash exposure while imposing obligations, and a large order can still leave capital tied up awaiting acceptance. This model is therefore an expectations test. Its task is to expose the combinations that would justify the financing valuation.

The valuable part is the system, but the bill includes the hardware

The strongest thesis joins two forms of reuse. Shared software can coordinate equipment; common engineering, procurement and tooling can make the next product less expensive to produce. The second mechanism can create substantial value even when no software license is sold. It should appear in contribution, development efficiency or investment per accepted unit, and it still has a physical cost.

There is evidence for software sold beyond an Anduril airframe: an $86 million, three-year mission-autonomy integration award from Special Operations Command in 2025, and a Lattice selection for NATO’s air-command data platform in July 2026. Neither announcement discloses a recurring software segment with known renewal rates or margins.

Software creates value through three economic roles. Software sold independently earns revenue. Software bundled with a platform can improve its price or contribution. Internal manufacturing software can reduce work in process or investment per unit. The last two benefits belong in the relevant operating assumptions; adding another software valuation afterward would count them twice.

Faster inspection might free cash from unfinished products and permit more output from existing tooling. Those improvements belong in the hardware cash flows. An independently paid software deployment earns revenue in its own right. Treating both as evidence of progress is sensible; valuing the operating improvement and then adding another software premium for the same improvement is not.

The customer wants interoperability

The Army’s Right to Integrate discussion describes open interfaces and modular architecture alongside Lattice. A customer that can combine vendors creates a market for a good integrator while preserving its ability to change suppliers. Anduril can benefit from being the system that works best in that setting. Its advantage must survive the customer’s continuing choice.

The Air Force’s June CCA awards put aircraft and mission-autonomy software into different competitions. Anduril won air-vehicle work and a competitive software option, alongside software awards to RTX Collins Aerospace and Shield AI. The service described a further competition for a primary provider by summer 2027. Winning the aircraft does not settle the software contest. ^S17

The same release links full licensing payment to performance and permits purchases from other vendors in its pool. Portability could let Anduril sell autonomy on competitors’ aircraft. It could also let competitors run software on Anduril hardware. The addressable market expands while customer captivity shrinks. ^S17

October’s NGC2 fielding expansion is important evidence of adoption, but its up-to-$1.8-billion value covers an integrated program, not software annual recurring revenue. Integration knowledge and field reliability can still create switching costs. The company has to keep earning them through development and support. ^S11

International growth can share the economics as well. The proposed EDGE-Anduril joint venture offers local production and sustainment, but a jointly owned operation is not wholly attributable parent revenue. More access to customers need not mean ownership of every dollar of profit. ^S15

More units do not guarantee the same growth in dollars

A May framework described a minimum of 3,000 surface-launched Barracuda-500M systems over three years. The announcement supplies neither unit prices nor program-level profits. Accepted deliveries and contribution after support costs will determine the economic result. ^S16

The Army’s $20 billion enterprise framework consolidates more than 120 procurement actions, reduces pass-through charges and sets volume discounts. It can lower selling and contracting friction, but some underlying business already existed. The buyer keeps competition for future programs. ^S5

I expect demand to favor replenishable systems, easier upgrades and mixed-fleet operation. That creates a price-volume tension. For an illustrative hardware product with unchanged mix, 20% more units at a 7% lower price means only 11.6% revenue growth. Treating greater affordability as a competitive advantage while holding both unit revenue and dollar growth unchanged would make the forecast internally inconsistent.

Steelfin’s common-component strategy could improve this equation. Flight testing is planned for 2027, however, so the announcement is a development milestone before accepted high-rate output. ^S9 Replenishment also differs from subscription revenue: initial inventory buildouts can give way to slower replacement cycles. Software upgrades may soften that pattern without eliminating it.

Ordering authority, funded work and a protected profit stream are different things. As a procurement comparison, FAR 52.249-2 permits covered fixed-price work to be terminated for convenience and describes settlement of work performed, allocable costs and reasonable profit on work done. It does not turn the full unperformed order into guaranteed margin. I have not verified that clause in the Anduril awards. Award-specific termination and acceptance terms are needed before a ceiling can support factory capital.

The base case needs a very large software business

A 22% consolidated operating margin requires an explanation of where Anduril earns bargaining power or saves cost. Lockheed Martin’s Missiles and Fire Control segment reported 14.1% for the first half of 2026, a comparison with different scope and mix with no implied ceiling for Anduril. The software-mix test below makes the explanation concrete instead of allowing the word autonomy to supply the margin. ^S12

Suppose hardware contributes 15% of revenue after product costs and development, independently licensed software contributes 50%, and corporate expense consumes 5% of total sales. The relationship is EBIT margin = hardware contribution × hardware share + software contribution × software share − corporate cost. A 22% EBIT margin requires 34.3% software sales; 28% requires 51.4%. These contribution rates and revenue mixes are author assumptions.

Translate the percentage into a business. At base-case 2036 sales of $50.7 billion, that requires $17.4 billion of software sales and $33.3 billion of hardware and related activity. At the $61 billion price’s reverse-DCF revenue hurdle, required software sales rise to $34.1 billion. The test solves for the software required to reach the assumed margin. Actual software sales remain undisclosed.

Manufacturing efficiency is a consequential alternative to a large independently licensed software business. At 20% hardware contribution, the required software share falls to 23.3%, or $11.8 billion, for the base margin. That route could support a strong company with less dependency on software rents. It still requires the economics to survive pricing pressure and full development expense. The two routes should meet in the same consolidated margin, with costs allocated once.

Scroll horizontally to inspect the figure →

Value is what the business can earn; liquidity is the cash it holds

Arsenal-2 combines $3.7 billion of announced private investment with a Navy contract of up to $2.9 billion tied to production outcomes. Initial operations are expected in 2030. Payment for delivered output is not automatically a reimbursement of factory investment; netting the two headlines would invent a financing benefit. ^S6

At the Arsenal-2 announcement, President Trump said taxpayers would receive 40% of the shipyard. Completed rights, priority, and consideration remain undisclosed. The announcement concerns the project, leaving the parent’s interest to a final waterfall.

The consolidated DCF must eventually be reconciled to project ownership and financing. A completed outside claim can be deducted at fair value or the project can be modeled on an attributable basis. Neither 40% of the whole parent nor a silent zero for the project claim is justified. The same discipline applies to customer funding: it can reduce capital the parent advances without becoming additional operating profit.

I phase the private investment over 2027–2031 at $0.60 billion, $0.85 billion, $0.90 billion, $0.85 billion and $0.50 billion. That is an author schedule, additional to ordinary investment in other capacity. The Navy ceiling does not reduce it.

The base reaches a $10.5 billion cumulative economic FCF deficit, $7.5 billion beyond hypothetical $3 billion liquidity. I charge stock compensation in that DCF. Treasury funding can be lower when shares replace cash pay, while new employee claims still cost owners value.

The compensation sensitivity shows why. Adding back assumed noncash compensation equal to 2% of sales reduces the cash-timing gap to $6.3 billion; at 4%, it falls to $5.4 billion. Enterprise value does not change. Stock compensation, withholding, repurchases, customer advances, restricted cash, debt service and liquidity reserves remain unknown. Neither percentage estimates Anduril’s disclosed pay mix.

Consider a hypothetical project worth $10 billion before $4 billion of project debt. If an outside party owns 40% of its residual equity on equal terms, its claim is $2.4 billion: 40% × ($10 billion − $4 billion). Deducting $4 billion as 40% of project enterprise value would misstate that equity claim. Deducting 40% of Anduril’s entire parent value would use the wrong business altogether.

The announced shipyard stake doesn’t define the hypothetical equal-ranking waterfall. The contract record must show what the project owns, what it owes and who receives the remaining cash. For comparison, FAR 52.232-16 illustrates how progress payments can finance incurred costs before acceptance and later be liquidated against invoices, with property rights and performance conditions attached. I do not assume Anduril receives those terms. They show why customer finance can improve cash timing without improving the underlying product margin.

Scroll horizontally to inspect the figure →

The base requires $50.7 billion of 2036 sales

Revenue rises to $6.9 billion in 2027, $22.2 billion in 2031 and $50.7 billion in 2036, approximately 28% annual growth over a decade. The margin-mix test gives that scale economic meaning; it does not turn an expectations model into inside knowledge of classified programs.

EBIT margins rise from a 20% loss in 2027 to 9% in 2031 and 22% in 2036. Ordinary capital spending declines from 6% to 3.5% of sales, before Arsenal-2. Operating working capital falls from 18% to 14% of sales, a favorable assumption about inventory, collections, supplier terms or customer funding.

In 2036, the model supports 12% sales growth with $1.27 billion of net plant and working-capital investment against $8.48 billion of after-tax operating profit, which means earlier factory spending, expensed development, and customer cash terms must keep later growth from requiring another proportional buildout of capital. I view that as an ambitious operating claim. Opening factories starts the test.

The five-year maturity transition then adds an explicit reinvestment adjustment as the model converges to sustainable returns. The downloadable forecast identifies that adjustment. The adjustment asks how much cash must stay in the business to make growth consistent with the assumed sustainable return on new investment. It is not a claim that a particular construction project has been identified. Likewise, the cash ratios don’t measure accounting return on capital: expensed research, asset vintages and the opening balance sheet are not reconstructed here.

Free cash flow is operating profit after cash tax, plus depreciation, less capital expenditure and the change in working capital. The model taxes positive EBIT at 24% and gives no immediate refund for losses. Undisclosed loss carryforwards could defer taxes, so no extra tax asset is added. Development and economic employee compensation stay inside EBIT.

The remaining-2026 stub is 78/365 of an annualized cash-flow proxy, approximately negative $0.53 billion. Its annual inputs are the reported $1.2 billion operating-loss projection, 2% depreciation, an assumed $1 billion ordinary capex budget and 18% working capital on the change in sales. The model distributes the annual proxy across that partial-year period. Earlier spending is not charged again.

The result is about $28 billion of operating value

I use 12% as an underwriting return hurdle. After 2036, growth fades over five years to 3%. Reinvestment converges to what sustains a 15% return on new invested capital in the base case.

Terminal cash flow equals next-year after-tax operating profit multiplied by one minus growth divided by return on new capital. At 3% growth and 15% returns, 20% of operating profit after tax must be reinvested. Faster growth enlarges the future profit pool, but also requires leaving more of that profit in the business to fund the expansion. About 66% of base enterprise value still comes from the terminal period after 2041.

The base result is $27.7 billion of enterprise value and $30.7 billion after hypothetical $3 billion net cash, before outside claims. A 10% discount rate raises that bridge to $46.7 billion; 14% lowers it to $21.1 billion. Those sensitivities vary the discount assumption. Cash and fair-valued outside claims change the bridge dollar for dollar.

The upside reaches $74.9 billion of 2036 sales, 28% margins and 20% mature returns on new capital: $67.4 billion of enterprise value and a $70.4 billion gross bridge. That combines faster growth, better margins and less capital intensity. It is not an independent probability estimate of three favorable events.

The downside grows to $17.1 billion of sales and 15% margins, yet continuation value is approximately zero after investment and losses. Its $2.9 billion gross bridge is not a cash-recovery floor. Project cancellation could improve outcomes; unavailable financing and senior claims could worsen them. Negative continuation value calls for a changed capital plan. Limited-liability shares can’t have a negative purchase price.

Scroll horizontally to inspect the figure →

Scroll horizontally to inspect the figure →

The larger risk is earning the margin too late

I stress the operations at the same 12% hurdle. Delaying the margin ramp two years cuts enterprise value from $27.7 billion to $19.2 billion. Phasing the mature margin down to 18% gives $20.3 billion. A slower earnings engine costs more than the 50% shipyard overrun, whose value impact is $1.3 billion.

Thirty additional collection days reduce value to $24.8 billion. That scenario builds extra receivables equal to sales × 30/365 beginning in 2027 and funds their growth permanently, including in the terminal period. It increases receivables and cash committed to customers while holding reported operating profit fixed. At 2031 base revenue alone, the extra balance is about $1.8 billion.

A 50% Arsenal-2 spending overrun lowers value to $26.4 billion. This conservative cost stress gives no extra depreciation or tax shield for the overrun. Its $1.3 billion present-value effect is much smaller than the $8.5 billion lost to a delayed margin ramp: a one-off construction problem and a slower earnings engine have different durations.

The combined operating stress gives $7.5 billion of enterprise value; independently calculated losses cannot simply be added. This combines the declared adverse assumptions. Its lesson is that sustained margin and collection timing affect many years of capital recovery. A one-off construction overrun matters, but the business can lose more value by producing for longer before it earns and collects its expected contribution.

Scroll horizontally to inspect the figure →

What would make $61 billion reasonable?

At base margins, capital productivity and discounting, the $61 billion gross equity reference requires $99.5 billion of 2036 sales: about 37% annual growth from the planning anchor. The $100 billion reference requires $162.9 billion, or 44% growth. Each annual growth factor changes, so earlier losses and investment scale with the larger business across the forecast.

Both tests assume $3 billion net cash and no minority deduction. Final outside claims would raise the required operating outcome. The companion stress solver can price those claims explicitly instead of assuming that a gross company valuation belongs entirely to common holders.

The bullish rebuttal must change margins, capital efficiency or the cost of financing growth in an identifiable way. Independent software wins could improve mix. Reusable tooling and customer financing could reduce capital the parent needs to advance. The upside supports $61 billion through such a combination; the base’s 22% margin already credits substantial improvement. Another platform premium outside those cash flows would add no new cash source.

Intrinsic value also differs from a venture return target. With no interim distributions, a 15% annual return over ten years and 20% cumulative ownership dilution require an exit equity value of about $308 billion from a $61 billion entry, or $506 billion from $100 billion. The equation is entry value × 1.15¹⁰ ÷ 0.80. These are exit-equity values required to hit the stated return after dilution.

Future investment is already charged through negative FCF. A blanket dilution haircut on top is not the right adjustment. New financing creates cash and a claim simultaneously; price, preferences and participation terms determine its effect on existing owners. The return illustration does not feed back into the DCF.

Scroll horizontally to inspect the figure →

A stronger bull case needs better margins and capital efficiency

Software and production could reinforce each other more strongly than the base allows. Common components can reduce repeated tooling, updates can keep equipment useful and third-party autonomy can reach another vendor’s airframe. Customer financing can also bridge production cash before delivery. The last mechanism changes timing and contractual obligations, while the others can change operating economics. Identifying each mechanism makes a stronger bull case than treating every award as recurring software.

The reverse model shows how much that matters. Phasing the mature EBIT margin from 22% to 26% lowers the 2036 revenue needed for the $61 billion bridge from $99.5 billion to $81.1 billion. The software and hardware contributions still need to support that margin; it is not an assumption that every added dollar has negligible cost. This is a more coherent bullish argument than applying a software multiple to consolidated sales.

There is also a financing-market argument. Applying a 10% discount rate to the entire upside operating case gives a gross equity bridge of $103.2 billion. That scenario combines $74.9 billion of 2036 sales, 28% margins, better capital efficiency and a lower required return. It demonstrates that $100 billion is arithmetically possible. It does not demonstrate that the combination is likely or that the smaller discount rate compensates for the execution risk.

I would revise the valuation as intermediate evidence supports those mechanisms. Repeat paid software use would support the licensing thesis; qualified output with less rework would support manufacturing reuse; executed financing and acceptance terms would support better capital recovery. No single milestone has to wait until 2036. Each has to change a line in the forecast before changing my price.

A public listing, a strategic investor or another high-priced private round can change liquidity and market price without, by itself, proving better underlying cash generation.

The evidence that would change my price

First, paid software adoption independent of Anduril hardware. Repeat licensing, third-party deployments and contribution after development would support a higher margin case. Another integrated contract with an AI label would not settle the question.

Second, accepted output and cash conversion. I would seek repeat shipments, first-pass yield, rework and warranty costs, inventory days, and capital required for additional qualified output. A growing order book is less persuasive when unfinished inventory grows faster and acceptance is delayed.

Third, I would reconcile attributable economics and contract capital. The shipyard’s final ownership and funding rights must reach the parent waterfall, while advances, progress payments, acceptance and termination provisions must reach the cash forecast. A customer can finance part of production and keep significant rights over it. That may still be attractive business; it is different from unrestricted capital or a guaranteed software annuity.

My expectation is that Anduril becomes more industrial as it becomes more important. The valuable outcome is a production and integration system that supplies capability more efficiently while the customer keeps architectural control. That is compatible with open competition and a strong company. It does not ensure that shareholders capture the entire benefit of the improvement.

I keep the base near $28 billion of enterprise value until paid contribution, capital recovery, and attributable rights justify more. At $61 billion I’d need evidence closer to the upside. Strategic importance earns Anduril serious study. Repeated cash left after serving the customer earns the investor’s premium.

Assumptions

• FY2025 revenue is company-reported; FY2026 revenue and operating loss are March investor projections reported by the press, not observed results or newly confirmed October guidance.
• The model uses a remaining-2026 stub, annual forecasts through 2036, a five-year maturity transition, and a reinvestment-funded terminal value. R&D and economic employee compensation remain in EBIT.
• Net cash and available liquidity are separate $3 billion assumptions. Zero minority-interest value means an unadjusted bridge, not evidence that outside claims are worthless.