The Twenty Minute VCShyam Sankar: The Broken Incentive Structure of How Governments Buy Defence | E1104
CHAPTERS
- 0:00 – 0:54
Fixing DoD procurement: fairness vs speed, and defense spending as % of GDP
The conversation opens with Sankar’s diagnosis of why modern defense procurement is slow and inefficient: excessive process intended to ensure fairness ends up delivering neither fairness nor speed. He also frames the stakes by contrasting today’s defense spending levels with Cold War-era spending as a share of GDP.
- •Process-heavy procurement undermines speed and real competition
- •EU procurement is even slower due to cross-state fairness requirements
- •Defense spending today is historically low vs the Cold War (as % of GDP)
- •Framing defense budgets as an “insurance premium” on GDP
- 0:54 – 3:26
Work ethic forged by family, adversity, and the ‘counterfactual’ mindset
Sankar traces his drive to his father’s example and his family’s experience of violence, displacement, and starting over in the U.S. He emphasizes gratitude, resilience, and a forward-looking focus on what to do next rather than dwelling on setbacks.
- •Father’s influence: obligation to family and making the most of opportunity
- •Family fled Nigeria after severe violence; resettled in the U.S. from scratch
- •“Counterfactual” perspective: remember how much worse it could have been
- •Productive response to hardship: stick together and keep moving forward
- 3:26 – 4:37
Hard work vs luck—and why agency matters more than attribution
Pressed on luck, Sankar argues it’s not productive to fixate on it because it’s not controllable. He believes the current era overemphasizes luck and underweights hard work and smart execution, even if both play a role in outcomes.
- •Luck may exist, but it’s not the variable you can control
- •Best mindset: assume good luck and focus on work and execution
- •Critique of modern narratives that over-credit luck
- •Success is a mix of factors, but effort is the actionable lever
- 4:37 – 8:56
Palantir’s early days: outsider engineers taking on classified national security problems
Sankar describes arriving in Silicon Valley via Xoom, then joining a tiny Palantir motivated by national security. With no clearances and little institutional knowledge, the team had to build by first principles in an environment with huge information asymmetry.
- •From Cornell to Stanford to Xoom; joined as an early employee
- •Motivation to work on national security shaped by 9/11 and personal events
- •Palantir was ~12 people; initially hired only engineers
- •Building for classified environments without clearances created major asymmetry
- 8:56 – 12:32
Inventing a new government sales model: refusing hourly services in favor of product licensing
Sankar contrasts the government’s default ‘cost-plus’ labor model with Palantir’s push for a product business model funded by venture capital. He explains why cost-plus discourages innovation and why transitioning government buyers to licensing/subscription software is difficult but essential.
- •Legacy model: cost-plus, hourly labor, audits, capped profits
- •Cost-plus incentives discourage private investment and innovation
- •Palantir strategy: build product, finance via VC, sell via licensing/subscription
- •Government procurement inertia still pulls vendors back to services
- 12:32 – 16:28
How decisions really get made: committees, misaligned buyer/user roles, and ‘land & expand’ in defense
This segment explains the byzantine internal mechanics of government buying, where users and buyers are often disconnected. Palantir’s expansion depends less on sales tactics and more on product-driven expansion—deeply embedding with operators to discover the next capability to build.
- •Decision making is committee-like; hard to map cleanly
- •Three essentials: requirement, resources (money), and contract
- •Buyer-user disconnect: win with users, then users influence buyers
- •Land & expand is product-surface expansion, not just more seats/licenses
- •Operational proximity matters: learn in the field, not from HQ
- 16:28 – 17:53
Why government buys hardware better than software—and why ‘procurement vs sustainment’ breaks down
Sankar argues the acquisition system is built for hardware programs with distinct phases (R&D, procurement, sustainment), which doesn’t fit modern software. Software is never “done,” and costs don’t drop the way procurement organizations expect them to.
- •Acquisition mindset optimized for hardware lifecycle phases
- •Government expects post-procurement costs to decline; software doesn’t behave that way
- •Modern reality: procurement and sustainment are continuous for software
- •Scaling in air-gapped/cleared environments requires heavy productized infrastructure
- 17:53 – 24:55
America’s defense industrial problem: WWII excellence, post–Cold War consolidation, and financialization
Sankar uses WWII mobilization as the benchmark, emphasizing founder-led dynamism and industrial scale-up. He then traces today’s dysfunction to post–Cold War consolidation (“The Last Supper”), shrinking the prime base and shifting incentives toward financial engineering over innovation.
- •WWII model: rapid industrial mobilization and founder-led innovation
- •18-month industrial ramp enabled U.S. production dominance
- •1993 “Last Supper” encouraged consolidation from 51 primes to ~5
- •Consolidation led to financialization: dividends, buybacks, M&A over invention
- •Result: weaker innovation incentives and mistrust between government and industry
- 24:55 – 29:23
Process, risk aversion, and the illusion of fairness: why procurement can’t move at battlefield speed
The discussion turns to how efforts to avoid protests and preserve public trust created bloated process and extreme risk aversion. Sankar argues this suppresses learning and constrains outcomes to mediocrity—especially compared to historical examples like Skunk Works shipping in months.
- •Too much process yields neither fairness nor speed
- •Europe’s fairness constraints often slow procurement further than the U.S.
- •Risk aversion prevents learning; mistakes are necessary for capability growth
- •Historical contrast: Kelly Johnson/Skunk Works shipped transformative aircraft fast
- •Over-constraint reduces variance but caps excellence—“mediocrity by design”
- 29:23 – 34:04
Pick winners (plural): aligning venture dynamics with defense acquisition and avoiding ‘zombie’ funding
Sankar explains why spreading small dollars across many startups kills the venture flywheel in defense tech. He advocates selecting a small number of winners at scale to create real outcomes that keep capital and talent committed to the ecosystem.
- •VC portfolios depend on a few big winners, not uniform mediocre outcomes
- •“Peanut butter spreading” creates zombies: neither dead nor able to scale
- •DoD claims it doesn’t pick winners, but every contract awards a winner
- •Scaling winners creates a virtuous cycle: returns → more capital → more innovation
- •Misalignment persists due to discomfort with concentrated outcomes
- 34:04 – 37:49
Defense budgets, new wars, and fast-changing tech: drones, attritable systems, and new logistics models
This chapter covers how recent conflicts accelerate learning about what works and reshapes what militaries should buy. Sankar highlights the move toward lower-cost, attritable, rapidly iterated systems—where decades-long sustainment models no longer apply.
- •Defense spending is ~2–3% of GDP vs ~5–6% at Cold War peaks
- •Tone of buyers stays serious; what changes is the rate of learning from battlefields
- •Shift toward “lower-cost, attritable” capabilities rather than perfect 30-year platforms
- •Ukraine demonstrates single-use drones: sustainment assumptions collapse
- •Rapid iteration changes supply chains, logistics, and procurement needs
- 37:49 – 42:23
If Sankar ran the DoD: competing programs, internal competition, and empowering the fighters’ vote
Asked what he’d do in charge, Sankar proposes multiple parallel programs pursuing the same capability to drive empirical learning and speed. He also argues for more internal competition within government and tighter feedback loops from combatant commanders to procurement decisions.
- •Use multiple competing programs per capability (historical precedent in missile programs)
- •Competition accelerates learning and reduces tech risk empirically
- •Current incentives reward PMs for cost/schedule compliance, not rapid innovation
- •Increase “internal competition” inside government, not just among vendors
- •Give combatant commanders stronger influence in what gets bought and in what mix
- 42:23 – 45:20
Elections vs procurement reality: why budgets and continuing resolutions matter more than politics
Sankar argues leadership changes matter less than people assume; the bigger determinant is whether funding is actually available. In the U.S., continuing resolutions can freeze new starts and stall the department’s desired initiatives.
- •Political leadership shifts have variable impact by country
- •UK: cabinet/treasury involvement makes politics more directly relevant
- •U.S.: Congress and the budget cycle dominate acquisition timing
- •Continuing resolutions prevent “new starts,” delaying modernization
- 45:20 – 57:31
From defense policy to company craft: ‘content over process’ and managing talent like an artist colony
The discussion pivots to Palantir’s internal philosophy: great outcomes come from substance (content), with process serving as support rather than dogma. Sankar extends the idea into hiring and management—design roles around asymmetric talent and create conditions for learning, autonomy, and long-arc performance.
- •Companies fail when they worship process instead of content (Jobs’ critique)
- •Process should be “medicine,” not an “opioid” that replaces thinking
- •Scale comes after substance; scaling itself may require repeated hard effort
- •Talent is asymmetric—better to craft roles around people than slot them into ladders
- •Manage creatives via autonomy, protection from distractions, and an environment for learning
- 57:31 – 1:06:12
Palantir’s reinventions and quick-fire: AI value accrual, tools, adoption gaps, and the next five years
Sankar reflects on the human cost of repeated reinvention—credibility resets and individuals must re-earn mastery. In the rapid-fire, he shares views on AI economics (value in apps and infrastructure, model squeeze), why LLMs need tools, and how geopolitics and AI could reshape institutions over the next five years.
- •Reinvention is hard because everyone effectively starts from zero again
- •At scale, reinvention requires broad communication but becomes a proud crucible
- •AI value: incumbents/app layer plus infrastructure; models get squeezed/commoditized
- •LLMs need tools to move beyond parametric knowledge and manipulate application state
- •Concern: focus on proof of value (production impact) vs impressive demos; geopolitics + AI will force institutional transformation
