Scale AI Loses to Smaller Startup in Bid for US Intel Work

Enabled Intelligence, a Falls Church, Virginia startup founded in 2020, won a seven-year contract to provide data labeling for AI/ML to the U.S. Department of Defense and intelligence community, beating larger rival Scale AI. The National Geospatial-Intelligence Agency disclosed the award after the government shutdown; the contract has a ceiling of $708 million and is the agency’s largest data-labeling effort to date. The decision underscores rising federal demand for labeled data for defense AI and represents a procurement setback for Scale while validating a smaller vendor in the defense AI supply chain.

Analysis

Market structure: The award validates a non‑incumbent supplier model and increases bargaining leverage for cleared, niche labeling vendors; expect a 5–15% bid‑premium for mid‑tier government IT/analytics contractors that can scale cleared labor in 6–18 months. Larger AI infra/software vendors see limited direct revenue upside from a $708m ceiling over 7 years (~$100m/year) but suffer reputational and competitive pressure — pricing power for labeling services will fragment, not consolidate. Cross‑asset: expect modest tightening in high‑grade defense credit spreads (5–30bps) if wider DoD services ramp; equity impacts should be concentrated in mid‑caps and cybersecurity names, with FX/commodities negligible.

Risk assessment: Tail risks include GAO/procurement protests, program declassification shifts, or a major data breach that could void contracts — any of which could wipe 30–70% of expected revenues for a small vendor. Immediate (days) volatility will track procurement headlines; short term (3–9 months) depends on task order issuance and workforce hiring; long term (2–5 years) hinges on DoD AI budget trajectory and classified dataset expansion. Hidden dependencies: cleared workforce, access to classified imagery, and subcontractor offshore restrictions — bottlenecks that can delay revenue by 6–12 months. Key catalysts: GAO protest timeline (30–120 days), FY26 DoD AI budget decisions (6–12 months), and first-year task order awards.

Trade implications: Favor selective long exposure to mid‑cap government IT/analytics and cybersecurity; avoid paying high multiples for pure‑play data labeling software absent recurring contracts. Pair trades: long cleared government services (BAH, CACI) vs short unprofitable AI infrastructure names that trade on narratives. Options: use 6–12 month call spreads to gain asymmetric exposure to winners while capping premium loss. Rotate 1–2% from general AI/growth into defense and cybersecurity over 2–6 weeks, scale after first task orders.

Contrarian angles: The market may overemphasize headline size; $100m/year is immaterial to large primes but material to mid‑caps — mispricing exists in mid‑cap contractors without visible labeling capability. Historical parallels: small vendors that won early classified DoD work (e.g., Palantir origins) can re‑rate, but many winners fail to scale due to clearance and ops constraints. Unintended consequences include stronger antitrust/procurement scrutiny and competitors accelerating low‑cost offshore labeling that compresses margins; those risks are actionable within 3–12 months.

Key Decisions for Investors

• Establish a 2.5% long position in Booz Allen Hamilton (BAH) within 2–6 weeks; target 12–18% upside over 6–12 months, set a hard stop at -8% and reduce position by 50% if no task orders or revenue guidance materializes within 9 months.
• Initiate a 1.5% long position in CACI International (CACI) now, add another 1.0% if management discloses >$50m of annualized classified AI services revenue or new task orders within 9 months; target +15% in 9–12 months, stop-loss -10%.
• Allocate 0.75% of portfolio to a 9‑month Palantir Technologies (PLTR) call spread (buy ~25‑delta call, sell ~55‑delta call) to capture upside from expanded DoD AI spending; close if PLTR rallies >25% or if a GAO protest is filed within 60 days.
• Reduce broad AI/growth ETF exposure by 2% (e.g., trim QQQ/ARKK holdings) and reallocate 1.5% to CrowdStrike (CRWD) and 0.5% to the iShares U.S. Aerospace & Defense ETF (ITA); review performance at 3 months and take profits on CRWD if it underperforms S&P by >10%.