Former Intel CEO on What Went Wrong, What's Next + Lovable CEO on the Real Promise of Vibe Coding

Pat Gelsinger joined Intel at 18 and spent 34 years watching the company evolve from technical leadership under Andy Grove, Gordon Moore, and Bob Noyce—where 15 of 20 executive staff held PhDs—to a business-led organization that prioritized spreadsheets over silicon. The turning point came when non-technical leaders began making billion-dollar technology decisions without understanding the underlying trends. In the five to six years before Gelsinger's return as CEO, Intel returned $100 billion to shareholders through dividends and buybacks while failing to build a single new factory in a decade.

The consequences were catastrophic: Intel passed on making chips for the iPhone, failed to invest in EUV lithography machines, and never developed a true foundry model. Meanwhile, Steve Jobs had quietly ported macOS to x86 architecture across four releases, preparing Apple to design its own silicon. Jensen Huang at Nvidia built a robust software stack (CUDA) that transformed graphics cards into general-purpose computing platforms, eventually dominating AI workloads. TSMC executed a visionary foundry strategy, becoming the manufacturing partner for the entire industry and ultimately producing 5× the wafer volume of Intel.

Gelsinger's diagnosis is unequivocal: «You don't do that through a spreadsheet. That's a lousy investment, right? Unless the technology trends make it the right investment.» He points to today's most successful tech companies—all led by deeply technical founders or engineers like Satya Nadella and Sundar Pichai. The lesson is universal: technology businesses must be run by technologists who understand how to make long-term bets that spreadsheets cannot justify but that the future will vindicate.

Gelsinger sees the AI infrastructure wave as fundamentally different from past technology bubbles because it is constrained by physics, not speculation. Global energy capacity is expanding at only 4–5% annually—and the U.S. managed just 1% growth over the past decade. Since no company will buy GPUs or build data centers without guaranteed power, energy availability acts as an upper bound on investment, preventing the kind of irrational exuberance that characterized the dot-com era.

Yet the potential value is nearly limitless. If a token represents a unit of intelligence, and intelligence drives better supply chains, more efficient logistics, superior financial analysis, and breakthrough science, then the economic return is «somewhat infinite,» especially in developed countries facing labor shortages. Gelsinger believes the industry is entering a multi-decade buildout, not a multi-year hype cycle. His stated objective is to make AI 10,000× more cost-effective—dropping the cost per token by five orders of magnitude to trigger Jevons' paradox and explode access to AI across every sector.

The next frontier is what he calls «the trinity of computing»: classical computing, AI computing, and quantum computing converging. When all three work together, industries will solve problems that cannot be computed today—from new materials and chemistry to curing cancer and lifting billions out of poverty. As Gelsinger puts it, «There has not been a time in human history where it's been better to be a technologist than the one we're in right now.»

Oika founded Lovable with a singular mission: empower humans to build great software, then help them run businesses around it. In 20 months, the platform has facilitated the creation of 50 million applications and now sees 1 million new projects built every week. Those apps collectively generate over 700 million visits per month, and the platform has reached $500 million in annualized revenue. Customers range from first-time non-technical founders to enterprise leaders who use Lovable to move faster than their internal engineering teams.

What sets Lovable apart from earlier no-code/low-code tools is that it produces genuinely good software: fast, secure, and architected according to best practices. The platform is opinionated—it handles payments, email integration, security scanning, search engine optimization, and data protection automatically. Customers who once would have spent $500,000 and six months on an internal tool now build it in 4–8 hours for the cost of a $50 monthly subscription. One enterprise customer replaced more than 10 internal tools with bespoke Lovable applications, saving over $1 million annually.

The product is evolving beyond building to operating. Lovable now offers an «AI co-founder» that works while customers sleep, analyzing business data and suggesting strategic optimizations. Under the hood, Lovable routes prompts to the best-suited frontier model (commercial or open-weight), then applies post-training and reinforcement learning to fix mistakes specific to its agent harness. About 60% of customers on the lowest subscription tier hit usage caps and pay overages—a sign they're extracting enormous value. As Oika puts it, «Many engineers don't look at the code, they don't write code anymore—and that means you don't need to be an engineer to create software.»