Dan Lewis knows waste when he sees it. He built a career spotting empty trucks rolling down highways with no cargo. Now he aims to tackle the empty cycles in the most expensive part of artificial intelligence.
Lewis left Microsoft this spring. The co-founder and former CEO of Convoy launched a stealth startup in May. Its goal sounds simple on the surface. Make AI models run cheaper, faster and at scale. Yet the ambition runs deeper. The supply chain for intelligence.
According to his LinkedIn profile, the new company constructs a computing platform spanning data centers, networking gear, specialized chips and real-time routing software for AI requests. The emphasis falls squarely on inference. That means executing trained models rather than creating them. Heavy workloads demand quick responses. Current systems often deliver neither speed nor affordability at volume. Lewis wants to change that equation.
“Our mission is to be the best stewards of power to make AI efficient, abundant, and affordable for this next era,” his profile states. Short statement. Large implications.
The announcement landed today in a GeekWire report. Lewis confirmed the move when contacted but offered few additional specifics. Too early, he said. Industry watchers already draw clear lines from his past successes and stumbles to this latest bet.
Convoy grew fast. The Seattle digital freight marketplace matched truckers with loads using machine learning. It priced shipments dynamically. It cut “deadhead” miles where rigs returned empty. Lewis and co-founder Grant Goodale scaled the startup to a $3.8 billion valuation. Then freight rates collapsed in 2023. The company shut down operations. Flexport bought the technology. Lewis advised there briefly before moving on.
His path before Convoy revealed an early focus on intelligence. Lewis studied cognitive science at Yale. He held an executive role at Wavii, the Seattle machine learning company Google acquired in 2013. At Amazon he shaped AI systems for product personalization. Even his first stint at Microsoft from 2008 to 2011 involved product management on Windows and Outlook.
He returned to Redmond in February 2025. This time as chief product officer for enterprise AI and Copilot efforts. He later became corporate vice president. Lewis helped companies construct AI agents and workflows. He launched an internal program called Camp AIR to speed AI-first teams. In a LinkedIn post at the time he described Microsoft’s direction with enthusiasm. The company raced toward a vision of giving millions of workers “AI superpowers.”
But the corporate role did not hold him long. By spring 2026 Lewis had departed. The new venture reflects lessons learned across those experiences. At the Seattle AI Startup Summit in April he shared hard-earned insights from Convoy without dwelling on its failure. “Be deliberate about culture,” he advised. Every organization develops one whether founders guide it or not.
He warned against assuming people read instructions. Convoy once built operations expecting drivers and staff to pore over notes. Most didn’t. Customers grew irritated. “Stop building a system that assumes people are going to read,” Lewis said. Design for reality instead. That principle could prove useful when routing millions of AI queries across global hardware.
Other lessons touched storytelling, data-driven decisions and hiring reluctantly. Name teams after problems, not solutions. Clarify expectations in both directions. None arrived easily. Convoy forced many of them into focus. Lewis spoke as someone who scaled a unicorn then watched it collapse amid market forces beyond control. Resilience shows in the new effort.
The timing feels pointed. Industry reports highlight surging demand for efficient AI inference as companies move beyond experimentation. A November 2025 Supply Chain Management Review analysis declared 2026 the age of the AI supply chain. Faster decisions, predictive planning and automation sit at the center. Yet power consumption and compute costs threaten to constrain progress.
Recent coverage echoes the pressure. Oliver Wyman’s EU Supply Chain Tech Report for 2026, produced with Prequel Ventures, details how executives industrialize artificial intelligence while partnering with startups for resilience and cost control. European firms in particular seek adaptable systems amid ongoing volatility. Lewis’s platform concept aligns with those needs even if his startup operates from Seattle.
Other startups chase adjacent opportunities. Y Combinator’s 2026 supply chain cohort includes companies applying AI to logistics optimization and decision layers atop warehouse and transport systems. Traction Technology’s review of hottest AI supply chain startups for 2026 spotlights firms using modeling, simulation and real-time analytics. None appear to target the foundational inference plumbing Lewis describes. That leaves room.
His approach draws direct parallels to Convoy’s model. The freight marketplace treated transportation capacity as a fluid resource to match against demand. Empty miles represented inefficiency to eliminate. The new startup treats compute, power and network bandwidth the same way. Route intelligence requests intelligently. Avoid idle hardware. Reduce energy waste. Make capability abundant rather than scarce.
But execution will test him. Building across data centers, chips and routing software requires deep hardware and systems expertise. Lewis brings product vision and AI experience. He lists himself as CEO and co-founder, suggesting partners fill technical gaps. Funding details remain undisclosed. Stealth mode buys time to assemble those pieces.
Challenges extend beyond technology. The original Convoy navigated brutal cyclical markets. AI infrastructure faces its own cycles. Chip shortages, energy constraints and regulatory scrutiny over power usage could complicate rollout. Lewis’s emphasis on stewardship of power signals awareness of environmental and cost pressures that grow more visible each quarter.
And yet his track record suggests persistence. After Convoy’s shutdown he joined Flexport, then Microsoft, then struck out again. Each move layered knowledge. Cognitive science background informs how humans interact with intelligent systems. Logistics experience shows how complex networks behave under pressure. Enterprise AI work revealed what large organizations actually need versus what technologists assume.
Industry conversation has shifted. No longer does talk center on training ever-larger models. Focus turns to serving those models to users efficiently and economically. Inference costs dominate budgets for many deployments. Speed determines whether AI agents feel magical or merely adequate. Lewis bets those problems represent an addressable market as large as the freight industry he once targeted.
Seattle provides fertile ground. The city hosts talent from Amazon, Microsoft and a cluster of logistics software firms. Lewis knows the players. His earlier lessons on culture and hiring could help attract the right team. “Hire deliberately and reluctantly,” he counseled at the summit. In an AI era some roles might disappear altogether. Better to ask whether the task needs a person or another model.
Details will emerge slowly. The company has no public name. No announced customers or partners. Lewis shared only that it remains early. Observers expect more within months as the team builds prototypes and seeks capital.
What seems clear is the continuity. Lewis spent years optimizing physical supply chains with software and machine learning. He now applies similar thinking to the invisible supply chains powering intelligence itself. Trucks, warehouses and routes become data centers, accelerators and inference graphs. Waste looks different. The impulse to eliminate it does not.
Success could reshape how organizations access AI capability. Lower costs and faster responses might accelerate adoption across industries still wary of high bills and latency. Failure would add another cautionary chapter to an already volatile sector. Either outcome will carry lessons Lewis will likely share again someday.
For now the entrepreneur moves quietly. From Yale classrooms to Google acquisitions, Amazon personalization, Convoy’s rise and fall, Microsoft’s AI push and this newest venture. The thread holds. Understand complex systems. Find the friction. Build mechanisms that route resources intelligently. Make abundance possible.
The trucks no longer roll empty on his watch. Soon the models might not idle either.
WebProNews is an iEntry Publication