SpaceX and Dell Score Massive Government Contracts
- 01Defense Spending Is Becoming a Defining Revenue Catalyst for Tech Companies
- 02The AI Infrastructure Stack
- 03Corporate AI ROI Is Hitting a Wall
- 04Defense Tech Is Globalizing and Attracting Non-Traditional Capital
- 05Physical-World AI Training Data Is the New Oil
SpaceX and Dell Score Massive Government Contracts
1. Key Themes
Defense Spending Is Becoming a Defining Revenue Catalyst for Tech Companies
Government contracts are now moving markets and minting billionaires overnight. SpaceX received $6.45 billion in new Space Force contracts — including $4.16 billion for "Golden Dome" missile defense satellites and $2.29 billion for a low-Earth-orbit communications network — just ahead of its expected IPO. Meanwhile, Dell's $9.7 billion Pentagon software contract sent shares surging more than 30%, lifting Michael Dell's net worth by roughly $34 billion and prompting the company to raise its full-year AI revenue forecast to $60 billion. The signal: proximity to U.S. defense priorities is now a legitimate moat and valuation catalyst.
"SpaceX received $6.45 billion in new U.S. Space Force contracts just ahead of its expected IPO." "Dell shares surged more than 30% after the company reported booming AI server sales and disclosed a $9.7 billion Pentagon software contract."
The AI Infrastructure Stack — Energy, Compute, and Connectivity — Is Attracting Massive Capital
Multiple funding rounds this week cluster around the same thesis: AI's physical backbone is undersupplied and overdemanded. Base Power is raising at a $12 billion valuation to manage home batteries and sell stored electricity back to the grid. Utilidata raised $40 million to orchestrate power use across data centers and grid-connected devices. Volt Harbor raised seed capital for modular energy storage for data centers. XCENA raised $135 million for memory-adjacent chips that reduce the cost and power demands of AI inference. Observable Space raised $90 million for laser communications infrastructure in orbit.
"Base Power...is reportedly in the market to raise a round at a $12 billion valuation." "XCENA...develops memory-adjacent chips designed to reduce the cost and power demands of AI inference." "Utilidata...orchestrates power use across data centers and grid-connected devices so utilities and AI data center operators can monitor capacity, manage distributed energy resources, and increase usable power."
Corporate AI ROI Is Hitting a Wall — and Cost Rationalization Is Underway
Despite the AI infrastructure boom, enterprise end-users are pulling back. Major companies saw token bills double or triple, or blew through annual AI budgets in months. Executives at Uber, Meta, Microsoft, Salesforce, and DoorDash are now steering employees toward cheaper tools, cutting overlapping products, and demanding clearer proof that AI spending is improving productivity.
"Corporate America is beginning to ration AI use after some companies saw token bills double or triple or blew through annual AI budgets in months, pushing executives at Uber, Meta, Microsoft, Salesforce, DoorDash, and others to steer employees toward cheaper tools, curb overlapping products, and demand clearer proof that AI spending is actually improving productivity."
Defense Tech Is Globalizing and Attracting Non-Traditional Capital
European and Asian defense startups are commanding billion-dollar valuations. Stark, a two-year-old Berlin startup making autonomous strike drones and loitering munitions, raised $350 million at a $2.9 billion valuation. Focused Energy, a Darmstadt-based laser fusion startup, raised a $240 million Series A backed by German utility RWE and the European Innovation Council. These are not marginal deals — they signal a structural shift in how non-U.S. capital is approaching defense and dual-use technology.
"Stark...raised a $350 million round at a $2.9 billion valuation." "Focused Energy...raised a $240 million Series A round. Investors included RWE, SPRIND, and European Innovation Council Fund."
Physical-World AI Training Data Is the New Oil
Robotics companies can't scrape the internet for embodied intelligence. AI training startups are now paying people — and offering free housecleaning — in exchange for video footage of everyday chores like washing dishes, folding laundry, and mopping floors. This represents an emerging, non-obvious labor market built entirely around generating proprietary physical-world data for robot training.
"AI training startups are paying people – and in some cases offering free housecleaning – in exchange for video footage of everyday chores like washing dishes, folding laundry, and mopping floors, as robotics companies hunt for real-world training data to help machines understand physical tasks that can't be scraped from the internet."
2. Contrarian Perspectives
AI Makes Developers Faster in Feel, But Slower in Fact — and Now They Won't Work Without It
The consensus is that AI coding tools are productivity multipliers. The METR research punctures that narrative twice over. First, developers who used AI in a 2025 study actually worked slower — they spent extra time finding and fixing errors, steering the AI, and waiting on it. Second, when METR attempted to replicate the study in 2026, developers refused to participate without AI access — making it impossible to measure improvement. The workforce has developed a dependency on a tool whose productivity benefit has not been rigorously established.
"Researchers measured how much time open source developers took to do tasks by hand versus with AI. While developers in that study reported that AI was making them more productive, they were shocked to learn it actually slowed them down." "Devs weren't willing to participate 'because they do not wish to work without AI' even just for the study."
SpaceX's IPO Is Being Bought on Fear, Not Fundamentals
Wall Street investors are openly acknowledging financial concerns — lower-than-expected space revenue and steep xAI losses — yet many say they'll buy the IPO anyway out of fear of missing out. This is a textbook FOMO-driven allocation dynamic. Retail investors are expected to play an unusually large role, amplifying the risk of valuation disconnected from underlying business performance.
"Wall Street investors are raising questions about SpaceX's finances, including lower-than-expected space-business revenue and steep xAI losses, but many say they may buy into the IPO anyway out of fear of missing what could be one of the largest public offerings ever, with retail investors expected to play an unusually large role."
Tesla's FSD Is More Theater Than Technology
Reuters reporting from former Tesla employees alleges the company used labor-intensive route mapping and training to make public robotaxi demos appear more capable than the underlying technology. AI trainers themselves observed the system struggling with stopping for school buses, avoiding emergency vehicles, and recognizing pedestrians and animals. If accurate, this suggests a significant gap between Tesla's public FSD narrative and actual autonomous capability.
"Tesla's own AI trainers have seen Full Self-Driving struggle with basic safety tasks, including stopping for school buses, avoiding emergency vehicles, and recognizing pedestrians and animals, while former employees say the company used labor-intensive route mapping and training to make public robotaxi demos look more capable than the technology really is."
3. Companies Identified
| Company | Description | Why Mentioned | Quote |
|---|---|---|---|
| SpaceX | Aerospace and defense company | Received $6.45B in U.S. Space Force contracts ahead of expected IPO; IPO financials under scrutiny | "$4.16 billion to build satellites for President Trump's 'Golden Dome' missile and air defense system and $2.29 billion for a low-Earth-orbit communications network." |
| Dell | Enterprise hardware and software | $9.7B Pentagon software contract; AI server boom sent shares up 30%+ | "Raised its full-year AI revenue forecast to $60 billion." |
| Base Power | Home battery installation and grid management | Raising at $12B valuation; Ribbit Capital as lead | "Installs and manages home batteries that provide backup power, lower residential electricity costs, and sell stored electricity back to the grid during peak demand." |
| Focused Energy | Laser-based nuclear fusion | $240M Series A; backed by RWE and European Innovation Council | "Developing laser-based fusion systems designed to generate clean power using deuterium-tritium fuel capsules compressed and ignited by lasers." |
| Garner Health | AI-driven healthcare navigation for employers | $100M Series E at $2.74B valuation; backed by Index, Kleiner, Sequoia, Founders Fund | "Uses healthcare data and AI to help employers steer employees toward higher-quality doctors, with financial incentives that lower out-of-pocket costs for patients." |
| Observable Space | Laser communications and space tracking | $90M round; backed by Lux Capital and RTX Ventures | "Makes laser communications ground stations, ground-based sensing systems, and in-space optical payloads for tracking objects, navigating spacecraft, and transmitting data from orbit." |
| Stark | Autonomous strike drones and loitering munitions | $350M at $2.9B valuation — two years old | "Makes autonomous strike drones and other defense systems, including loitering munitions designed to identify and strike targets." |
| XCENA | Memory-adjacent AI inference chips | $135M Series B at $570M valuation; South Korean | "Develops memory-adjacent chips designed to reduce the cost and power demands of AI inference." |
| Picogrid | Military sensor and command-and-control connectivity | $45M Series A led by Bessemer | "Develops hardware and software for connecting military sensors, autonomous systems, and command-and-control tools." |
| Utilidata | Grid and data center power orchestration | $40M Series C | "Orchestrates power use across data centers and grid-connected devices so utilities and AI data center operators can monitor capacity, manage distributed energy resources, and increase usable power." |
| MokN | Identity-based attack detection using decoys | $15M Series A led by GV with Datadog investing | "Protects organizations from identity-based attacks by using decoy corporate access points to expose stolen credentials before attackers can exploit them." |
| Modiqo | AI agent workflow automation | $3M pre-seed; turns agent executions into repeatable workflows | "Turns successful AI agent executions into repeatable workflows so enterprises can reduce token usage, improve reliability, and avoid rebuilding agent processes when models or APIs change." |
| Cognition / Devin | AI coding agent | Commits 89% of its own engineers' code | "Devin already commits 89% of its own engineers' code." |
| Tesla | Autonomous vehicles | FSD capability questioned by former employees and Reuters investigation | "Former employees say the company used labor-intensive route mapping and training to make public robotaxi demos look more capable than the technology really is." |
| METR | AI research lab | Published research showing AI slows developers and found devs refuse to work without AI | "Most developers won't work, even on a limited number of tasks, without AI anymore." |
4. People Identified
| Person | Description | Why Mentioned | Quote |
|---|---|---|---|
| Peter Thiel | Billionaire investor and PayPal co-founder | Building a foothold in Argentina as a "Plan B" amid concerns about U.S. taxes, politics, nuclear war, and runaway AI | "Buying a mansion in Buenos Aires, meeting with President Javier Milei and other officials, and temporarily relocating his family there." |
| Michael Dell | CEO and founder of Dell Technologies | Net worth increased ~$34B in a single day following Dell's earnings and Pentagon contract announcement | "Lifting Michael Dell's net worth by roughly $34 billion." |
| Scott Wu | CEO of Cognition | Claims Devin is meant to augment, not replace, engineers — even as it writes 89% of their code | "Devin, the company's AI coding agent, is meant to augment programmers rather than replace them, even though Devin already commits 89% of its own engineers' code." |
5. Operating Insights
Track AI Spend Like a P&L Line — Not a Utility Bill
The wave of enterprise AI cost rationalization is a direct operating signal: unmanaged AI adoption is a budget risk. The companies now walking back AI spend — Uber, Meta, Microsoft, Salesforce, DoorDash — went in without guardrails and got burned. Operators should instrument token usage by team and use case before scaling, and require teams to demonstrate measurable productivity outcomes before renewing or expanding AI tool subscriptions.
"Some companies saw token bills double or triple or blew through annual AI budgets in months, pushing executives...to steer employees toward cheaper tools, curb overlapping products, and demand clearer proof that AI spending is actually improving productivity."
Don't Assume AI Coding Tools Are Increasing Your Team's Output
The METR findings should prompt engineering leaders to audit actual throughput — not perceived productivity. Developers feel faster but may be slower in practice. Before headcount decisions are made on the assumption that AI has multiplied developer capacity, measure cycle time and defect rates empirically. The risk is making staffing cuts based on productivity gains that haven't actually materialized.
"While developers in that study reported that AI was making them more productive, they were shocked to learn it actually slowed them down. Sure, it generated code faster, but then they spent extra time finding and fixing errors, steering the AI and waiting on it to complete tasks."
6. Overlooked Insights
The Robotics Training Data Market Is an Emerging Paid Labor Ecosystem
Buried in the "Essential Reads" section is a signal worth isolating: the hunt for physical-world training data is creating a new category of paid human labor — one that doesn't require technical skills, only domestic activity. This is structurally different from text-based data labeling. Companies willing to build proprietary pipelines for this kind of embodied data collection could create durable competitive advantages in humanoid robotics, and the data itself may become a valuable asset class.
"AI training startups are paying people – and in some cases offering free housecleaning – in exchange for video footage of everyday chores like washing dishes, folding laundry, and mopping floors, as robotics companies hunt for real-world training data to help machines understand physical tasks that can't be scraped from the internet."
Modiqo Points to an Emerging "Agent Reliability" Infrastructure Layer
Modiqo — a brand-new, pre-revenue, pre-seed company — is solving a problem that will only grow: successful AI agent runs are ephemeral and non-repeatable when models or APIs change. The company's approach of converting successful agent executions into durable, reusable workflows suggests an emerging infrastructure category around agent reliability and continuity that sits above the model layer and below the application layer. This is early, but the problem it addresses compounds as enterprises deploy more agents.
"Turns successful AI agent executions into repeatable workflows so enterprises can reduce token usage, improve reliability, and avoid rebuilding agent processes when models or APIs change."