Leidos earns mid-teens ROIC on durable DoD/intel work, but at $149.23 the market pays 103x earnings and 5.1x intrinsic value for a low-margin services book. Wait for a much lower price.

Plain English

Leidos is the company the U.S. government hires to run its computers, build its spy software, give veterans medical exams, and scan luggage at airports. It is one of the biggest such companies, with a workforce of cleared engineers that competitors cannot easily copy. The business is steady but boring. Today people are paying about five dollars for every one dollar of business value, because they are excited about defense and AI. That is a bad price for a fine business. Wait until it is cheap.

Thesis

Leidos Holdings is the largest pure-play federal IT and mission-services contractor in the United States, with roughly 90% of revenue coming from the U.S. government — Defense, Intelligence, Civil (FAA, NIH, CDC, DHS), and Health (VA disability exams). The business compounds slowly: long-cycle cost-plus and fixed-price contracts, very high re-compete win rates, a $40B+ backlog, and a reputation for executing on classified work that almost no commercial firm can replicate. Ten-year average ROIC is 14.71%, which is genuinely good for a labor-arbitrage services business, and net debt to EBITDA is -1.14 (i.e., net cash on a leverage-adjusted basis), so the balance sheet is a fortress.

The trouble is price. The scorecard shows P/E TTM of 103.63 against a 10-year average of 37.93, FCF conversion over the last five years of 0.0% (NOPAT has actually declined per scorer notes), and a reverse-DCF that requires 17.88% growth in perpetuity to justify today's quote. Intrinsic value comes in at $29.24 base / $37.10 high. At $149.23, price-to-IV is 5.10x. Even granting that owner-earnings TTM ran $296M and the IV calc may be punitive on a one-year cyclical low, the gap is so wide that no reasonable margin of safety exists today.

This is a Buffett-Munger "right business, wrong price" file. I would happily own LDOS at roughly bull-case IV ($37) or below; below $30 it becomes interesting. Above $50 the math stops working in any conservative framing. Today's price is a no-touch.

Moat

Leidos earns the kind of moat that Buffett describes when he distinguishes "businesses that we thoroughly understand, with durable advantages" [1] from glamorous but fragile compounders. The moat here is real but narrow, and it lives almost entirely in a single source: intangible regulatory and security barriers, with secondary cost-advantage and switching-cost reinforcement.

Intangibles (regulatory / security clearances). This is the dominant moat. Leidos employs tens of thousands of cleared engineers (Secret, Top Secret, TS/SCI with polygraph). The pipeline to clear a new hire runs 12-24 months and is gated by the U.S. government, not by money. A $10B competitor cannot simply hire away the cleared workforce because (a) clearances do not transfer to non-incumbents without re-investigation and a sponsoring contract, (b) non-compete and recompete rebadging rules slow movement, and (c) past-performance scoring on federal RFPs systematically favors incumbents who have already executed similar classified scope. This is the same kind of structural barrier Munger describes when he talks about businesses that benefit from regulatory inertia. Erosion risk is low over 5-10 years; it becomes meaningful only if the U.S. government insources work or fundamentally reforms FAR-based procurement.

Switching costs. Once Leidos is embedded in a program — say, the FAA's en-route automation, the VA's disability-exam workflow, or DISA's enterprise networking — the switching cost is enormous. Mission systems are layered on years of code, ATOs (Authority to Operate), and tribal knowledge. A re-compete loss usually means a 12-24 month phased transition with continuity-of-service risk that program managers detest. Re-compete win rates across the industry sit in the 80-90% range, and Leidos performs at or above that bar. Stress test: a $10B challenger with 5 years to dislodge Leidos from a flagship program (e.g., DES, the VA exam contract worth more than $30B over its life) would still face incumbent past-performance advantages and transition risk.

Cost advantages. Modest. Scale lets Leidos amortize bid & proposal cost, indirect rate pools, and shared services across $16B+ of revenue. But the underlying business is people-on-contract; gross margins run mid-teens because government cost-accounting rules (DCAA, FAR Part 31) cap allowable indirect rates and squeeze margin uplift. There is no Costco-style cost-per-unit chasm versus a focused competitor like SAIC, CACI, Booz Allen, or a Lockheed/RTX services arm.

Pricing power. Weak. Pricing on cost-plus work is reimbursed at audited rates with a fee cap; on fixed-price work, the contractor takes execution risk for a fixed margin. Best-and-final pricing is competitive on every recompete. The classic Buffett test — "can you raise prices 10% without losing customers?" — fails here. What Leidos has instead is volume durability: the customer needs the work done and the alternatives are few.

Network effects. None of consequence. Programs are bilateral customer relationships; there is no platform value that grows with users.

Competitor stress test. If you handed a competitor $10B and 5 years and told them to take 25% of Leidos's revenue, they would fail. They could win one or two flagship recompetes by aggressive pricing — and indeed Leidos has lost programs that way — but they could not replicate the cleared-workforce footprint, the past-performance citations, or the depth of agency relationships across DoD, IC, FAA, VA, and CMS in five years. They could in fifteen.

Erosion risks. (1) Insourcing — the government periodically converts contractor positions to civil-service. (2) Procurement reform that elevates small businesses or commercial cloud-native vendors (Palantir, AWS, Microsoft, Google) on traditional services scope. (3) Generative AI compressing labor hours on requirements analysis, software sustainment, and case adjudication — a real medium-term risk to seat-based services revenue.

Moat verdict: NARROW.

Management

Leidos's capital allocation track record is competent rather than exceptional, and the scorecard reflects this — capital allocation scores 15 out of a possible 25 on the rubric. Five-choice review:

1. Reinvest in the business. The business is asset-light; capex runs roughly 1% of revenue. Reinvestment shows up primarily through internally-funded R&D and IRAD on classified pursuits, plus capability acquisitions tucked into bid teams. There is no large organic reinvestment opportunity at attractive incremental returns — this is intrinsic to government services. The scorer flags ROIIC as not meaningful because NOPAT declined, which is a yellow flag.

2. Acquire other businesses. Leidos's modern shape was set by the 2016 reverse-Morris-trust merger with Lockheed Martin's IS&GS business, which roughly doubled revenue and was, by most measures, well-executed. Subsequent deals — Dynetics (2019, defense systems), L3 Security Detection (2020, airport scanners), Gibbs & Cox (2021, naval architecture), 1901 Group, and others — have been bolt-on size and mixed in returns. Security Detection in particular underperformed expectations. The pattern is reasonable: pay 8-12x EBITDA for capabilities adjacent to existing customer relationships. No major value destruction, no spectacular wins.

3. Pay down debt. The company has steadily de-levered. Net debt to EBITDA at -1.14 indicates a balance sheet that is, on a leverage-adjusted basis, in net-cash territory. This is conservative for the industry and unambiguously good. Score this choice an A.

4. Buy back stock. Leidos has been a regular but unspectacular buyer. Diluted share count has grown 7.17% over 10 years (per scorecard), meaning buybacks have not even offset stock-based compensation in aggregate over the cycle. There is no clear evidence that management has timed buybacks with reference to intrinsic value; in particular, repurchases continued at prices well above current IV ranges. Buffett's standard — "buy back only when shares trade meaningfully below conservative IV" — is not visibly the operating principle. This is the weakest item in the allocation review.

5. Dividends. Leidos pays a modest dividend (yield around 1%). Coverage is comfortable; the policy is unremarkable and rational for the cash-generative profile.

Communication quality. Investor communications are clear, segment disclosure is good, and management has not over-promised on margin trajectory in the way some peers (notably during the 2020-2022 SES integration period) did. CEO Tom Bell, who took the role in 2024, has been more disciplined than his predecessor on margin commitments and has emphasized portfolio focus on Defense, Intelligence, Civil, and Health rather than chasing every adjacency. Early days, but the communication style is closer to what Buffett describes as "high integrity leaders who understand their customers and act like owners" [1] than to typical defense-services CEO bromides.

Owner orientation. Insider ownership is low in absolute dollar terms — typical of a federal services public company — and compensation is heavily PSU/RSU based with TSR and adjusted-EBITDA metrics. This is industry-standard but not Buffett-grade owner alignment.

Net. A balance sheet built like a fortress, an M&A track record that is fine, buybacks that have not visibly offset dilution, and a credible new CEO. Nothing on this list is a red flag, but nothing pattern-matches to capital-allocation greatness either. Note that a 14.71% ten-year ROIC against a share-count that has grown 7.17% over the same window means underlying per-share economics have compounded more slowly than headline returns suggest.

Capital allocator: B-.

Industry

Federal IT and mission services is a structurally average industry with pockets of excellence. Porter's Five Forces:

Buyer power: HIGH and structural. The U.S. government is a near-monopsony customer. The Federal Acquisition Regulation (FAR), Defense Federal Acquisition Regulation Supplement (DFARS), and DCAA cost audits constrain pricing, fee, and indirect-rate recovery. Lowest-Price-Technically-Acceptable (LPTA) competitions, when used, compress margin to nothing. Best-Value tradeoff is more common in classified work and partially offsets buyer power, but the customer always holds the pen.

Supplier power: LOW to MODERATE. Inputs are people, software licenses, and OEM hardware (Cisco, Dell, Microsoft, AWS, Palantir). The labor market for cleared engineers is genuinely tight — that is supplier power exercised by employees, manifesting as wage inflation that has run ahead of contract escalators in 2022-2024. AWS/Microsoft/Google in classified clouds (C2S, JWCC) are increasingly necessary partners with real bargaining position.

Threat of new entrants: LOW for legacy scope, RISING for tech-forward scope. Building a cleared workforce and past-performance bench takes 15+ years; this is a near-insurmountable barrier for traditional services scope. But Anduril, Palantir, Shield AI, and AWS/Azure/Google Public Sector are entering with software-defined offerings that bypass the traditional staffing model. The procurement system is starting (slowly) to accommodate them via OTA, SBIR, and CSO vehicles. Over a decade, this is a meaningful share-shift risk.

Threat of substitutes: MODERATE and rising. Generative AI is a credible substitute for hours of requirements analysis, software sustainment, case adjudication (relevant to Leidos's VA exam business), and tier-1 IT support. Government insourcing is a perennial substitute. Commercial cloud is a partial substitute for managed-services scope. Over five years these compress hours billed on legacy programs.

Rivalry among existing competitors: HIGH on price, MODERATE on capability. The competitive set — SAIC, CACI, Booz Allen Hamilton, ManTech (now part of Carlyle), Peraton, GDIT, Leidos itself, plus services arms of Lockheed, RTX, Northrop, and Boeing — is concentrated but fragmented enough that every recompete is contested. Win rates on incumbent recompetes are high (~85%) but new-business win rates run 20-30%, and pricing pressure on best-and-final is constant. Margins across the industry cluster in a tight 8-11% adjusted EBITDA band; nobody escapes the gravity of cost-plus pricing.

Value pool location and trajectory. The value pool is slowly migrating from labor-hours-billed to software-and-data products. Vendors who can productize (Palantir is the clearest example) capture far more value per dollar of customer spend than those who staff to a contract. Leidos has tried to move in this direction (Trusted Mission AI, SE Core, digital modernization), but the bulk of revenue and profit remains in traditional services. Government IT spend is non-cyclical and grows at GDP-plus, but the share of that pool addressable by traditional services is shrinking at the margin.

Industry Verdict: Average.

Inversion

I am now playing the short-seller. The bull case for Leidos at $149 requires almost everything to go right; the bear case requires only a few things to go ordinarily.

1. The single event that kills this. A serious procurement-reform push under a cost-cutting administration — DOGE-style or otherwise — that takes a hatchet to federal services contractor headcount, cancels or de-scopes a flagship program (the VA disability-exam contract is the obvious target, given its political visibility and size), and accelerates the migration of low-complexity services to commercial cloud and AI-augmented tooling. One re-compete loss on a top-five program (DES, NGEN, GSM-O II, T-Cloud, FAA en-route) combined with a fixed-price program write-down would be enough to crater the multiple. The combination is not a tail event — Leidos has taken multi-hundred-million-dollar charges on Sentinel-class fixed-price programs before.

2. Why the moat is narrower than bulls think. The bull narrative is "cleared workforce, deep agency relationships, irreplaceable." The reality: every major program has 2-4 credible competitors with broadly equivalent clearance footprints. Past-performance is not a moat against an incumbent who is also capable (SAIC, CACI, BAH, GDIT all are). Re-compete win rates of 85% mean a ~15% loss rate at every recompete; over a 10-year window a $1B program has roughly a 28% probability of being lost to a competitor at least once. The "moat" is more like a turnstile that mostly works. Crucially, the moat does not produce pricing power — gross margins are mid-teens and indirect rates are audited. A moat that does not produce pricing power is a moat against extinction, not a moat that compounds value.

3. Why management is worse than it appears. Share count has grown 7.17% over 10 years despite billions in buybacks — a strong signal that buybacks were used at prices that did not protect per-share economics, and that stock-based compensation has been generous. The 2016 IS&GS deal was good; the post-2018 bolt-on M&A program (Dynetics, Security Detection, Gibbs & Cox) has produced unimpressive incremental returns and at least one impairment. The pattern is rational on each individual deal but does not aggregate to capital-allocation excellence. The 14.71% ten-year ROIC has not translated into compounding owner-earnings — owner-earnings TTM is $296M against a market cap above $20B; that's a ~1.4% owner-earnings yield, and the trend is flat-to-down per scorer notes.

4. What bulls are extrapolating that won't hold. Bulls extrapolate (a) recent revenue growth above 7%, (b) margin expansion to 11%+, (c) sustained recompete win rates, and (d) AI as a tailwind to the contractor model. Each is contestable. (a) Revenue growth is partly headcount + partly classified ramp; the classified ramp is volatile and partly already in numbers. (b) Margin expansion in a cost-plus / FAR-audited business is hard to sustain — every dollar of margin expansion is renegotiated at the next recompete. (c) Win rates have already been very high; mean-reversion is the base case. (d) AI is far more likely to compress billable hours on services contracts than to create a new revenue line for a non-product company; Palantir and Anduril capture the AI value pool, not the staff-augmentation primes.

5. Valuation trap. This is the heart of the bear case. P/E TTM is 103.63 against a 10-year average of 37.93. Reverse-DCF requires 17.88% growth to justify $149.23. The scorecard shows P/IV at 5.10x. If LDOS simply re-rates to its own 10-year average P/E (37.93x), and earnings stay flat, the stock falls roughly 63%. If earnings normalize toward the FCF conversion the scorer flags (0% over 5 years means TTM earnings are not a real owner-earnings number), the multiple looks even more punitive. The base IV of $29.24 and high IV of $37.10 are uncomfortable to defend — they imply 75-80% downside — but they are produced by the same deterministic methodology used across the scorecard universe and they are anchored to owner-earnings, not GAAP earnings.

The market is pricing this defense-services contractor as if it were Palantir. It is not Palantir. It is a competent, disciplined, capped-margin services business whose long-run earnings power probably grows at GDP-plus-3% and trades, in normal times, at 18-20x earnings.

If I am right, the stock could be worth $40-$60 within 2-3 years.

Lollapalooza Bias Check

Several biases are active in me as I write this and I want to name them honestly.

Anchoring. I am anchored to the scorecard's IV range of $29.24-$37.10 because the methodology is deterministic and presented as ground truth. The "do not redo the math" instruction reinforces this anchor. The risk is that the IV calc may be punitive on a one-year cyclical low in NOPAT and may not reflect normalized earnings power. I have noted this in the thesis but I have not let it override the IV-based recommendation, which I think is the correct discipline given a P/IV of 5.10x — even doubling the IV does not close the gap.

Authority bias / social proof. Federal services contractors trade at multiples around 18-22x earnings as a peer group. Leidos at 103x stands out as an outlier even within its own peer set, and seeing peers trade lower makes me more confident in the bear case. This is partly legitimate (peer multiples are real information) and partly social proof. I have tried to ground the bear case in business mechanics rather than "peers trade lower."

Recency bias. The narrative around "AI demand" and "great-power competition" is fresh and compelling. Recent quarterly results have been good. It would be easy to extrapolate. I am consciously discounting recency on growth and on margin expansion.

Confirmation bias. Once I noticed P/IV of 5.10x and P/E of 103, I was looking for reasons the stock should fall. I have tried to steelman the bull case in the moat section: the cleared-workforce barrier and the procurement principal-agent dynamics are genuinely strong tailwinds for incumbency. They are not strong enough at this price.

Deprival super-reaction. Federal services has been a quiet compounder for years and there is a real "don't miss the next leg" pull. I am consciously ignoring this — Buffett's discipline is that compounding only works if you buy at a price that gives a margin of safety. There is no margin of safety here. There will be other chances.

Commitment / consistency. Once I committed to "Avoid" I felt pressure to make every section align. I have tried to keep the moat and management sections genuinely fair (NARROW moat, B- management) rather than dragging them down to fit the verdict.

The biases that worry me most: anchoring to the deterministic IV, and confirmation bias. Both push toward "Avoid." The biases that push the other way (recency, deprival) are weaker for me on this name. I think the verdict is robust to bias correction in either direction.

10-Year Outlook

Ten years out, in 2036, I expect Leidos's business to look fundamentally similar but smaller in share-of-customer-spend than today. Same customers (DoD, Intel Community, FAA, VA, CMS, DHS) — these institutions are durable. Same business model — staff-augmentation, mission systems, fixed-price programs — but with two important shifts.

Customer base. Roughly the same. Federal IT spend grows at GDP-plus and is as close to non-cyclical as commercial demand gets. Leidos's customer concentration is a feature, not a bug, of the business model. I have HIGH confidence that the customers exist in 2036.

Profit per customer. Probably modestly higher in nominal terms but flat to declining in real terms. The fundamental driver: AI and software-defined services compress billable hours per program. Leidos will partially offset this by climbing the value chain (more software, more managed services, more outcome-based contracts) but the headwind is real and the contractor business model is on the wrong side of it. MEDIUM confidence on this point — could go either way depending on how aggressively procurement reform proceeds.

Moat. I expect the moat to be narrower in 2036 than today, not wider. The cleared-workforce barrier is durable, but the addressable scope of work behind that barrier is shrinking as AI-and-cloud takes the easier tier of work. The remaining work is harder and more profitable per hour, but there is less of it. MEDIUM confidence.

Single biggest threat. The procurement-reform-plus-AI combination, executed by an administration with a mandate to cut federal headcount. This is not a tail risk; it is the modal scenario over a 10-year window.

Confidence. The business will exist. The economics will be capped. The current price has nothing to do with either of those facts.

CONFIDENCE: medium

Position Guidance

• Recommendation: Avoid
• Conviction: High
• Target buy price: $40 (roughly bull-case IV; meaningful margin of safety begins below $35)
• Target trim price: $55 (above bull-case IV plus a normalization premium; current $149 is far above any defensible IV)
• Position sizing: 0% today. If price reaches $40 in a market-wide drawdown without business deterioration, build to a 2-3% position. If it reaches $30 with the business intact, scale to 5%. Cap at 5% given NARROW moat and capped industry economics.