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

Pat Gelsinger's Diagnosis of Intel's Decline

• Identified a critical strategic error starting in the 1990s: Intel transitioned from being run by deeply technical leaders (e.g., Andy Grove, Gordon Moore, Bob Noyce) to being managed by business-focused "bean counters" and finance professionals.
• Noted that upon becoming CEO in 2001, he was the first technical leader in 15 years, highlighting a 15-year gap where non-technical leadership dominated decision-making for a core technology business.
• Criticized a decade of financial prioritization where Intel distributed approximately $100 billion to shareholders via dividends and stock buybacks between 2016 and 2021, capital that could have funded new fabrication facilities and EUV machine acquisition.
• Highlighted that Intel failed to build a new factory for ten years prior to his return, missing the critical window to acquire EUV lithography machines and compete on manufacturing scale.
• Recalled passing on the opportunity to manufacture chips for the iPhone, a decision that allowed competitors to capture a dominant market position.
• Observed that Apple's transition to its own silicon (Apple Silicon) was driven by Steve Jobs' realization that Intel could no longer meet the demands for smaller, lower-power chips, a process Jobs prepared for silently over four OS releases before executing.
• Contrasted Intel's culture with Apple's "Skating to where the puck is going" approach, noting Apple's willingness to invest in small, internal acquisitions (like P.Semi) to build core semiconductor competence over time.

Competitive Landscape and NVIDIA's Rise

• Described NVIDIA's rise not as luck, but as the result of Jensen Huang's consistent iteration on hardware and the development of the CUDA software stack, which created a robust ecosystem for high-performance computing.
• Noted that Intel's "Larabee" project, which attempted a similar general-purpose GPU approach on x86 architecture, was cancelled by company leadership shortly after Gelsinger's departure, illustrating the cost of missed strategic pivots.
• Explained that NVIDIA's success in AI and cryptocurrency mining was an unforeseen byproduct of their technology becoming the optimal solution for general-purpose computing workloads, rather than a pre-planned strategy.
• Identified TSMC as a primary disruptor, noting they launched a foundry vision in the early 2000s while Intel remained an Integrated Device Manufacturer (IDM) unwilling to offer its manufacturing process to third parties.
• Cited that when Gelsinger returned to Intel, TSMC was producing five times the number of wafers as Intel, a ratio that has since widened to seven-to-one.
• Stated that Intel's new strategy involves becoming a foundry for other companies, aiming to close the gap with TSMC, though the lead remains significant.

Geopolitical Risks and the Taiwan Semiconductor Ecosystem

• Warned that Taiwan possesses less than three weeks of energy reserves, creating a catastrophic risk where a blockade could cause a "brownout" that shuts down fabs for 90 days, with economic impacts exceeding the Great Depression.
• Highlighted that China has blockaded the Taiwan Strait seven times in the last four years, with military exercises indicating clear intent to resolve the conflict, potentially by 2027.
• Asserted that the US is currently building only 18% of leading-edge chips (up from 12% when Gelsinger returned in 2001), indicating a long road to self-sufficiency despite the CHIPS Act.
• Predicted that by 2030, the industry will have established meaningful resilience, though the current supply chain remains fragile.

AI Market Dynamics and the "Bubble" Argument

• Acknowledged high valuations in the AI sector but argued that energy capacity acts as a natural "circuit breaker" or upper bound, preventing a bubble as extreme as 1999's dot-com crash.
• Stated that data center growth is constrained by energy grid expansion, which in the US has only grown 4-5% annually recently, contrasting with previous decades of 1% growth.
• Expressed optimism for a multi-decade build-out of AI, aiming to reduce the cost per token by five orders of magnitude to achieve Jevons Law (where lower costs drive exponentially higher consumption).
• Predicted that quantum computing will yield meaningful results across multiple industries (chemistry, biology, logistics) by 2030, following the proven ability to error-correct qubits across various modalities (trapped ions, photonics, spin).
• Forecasted that quantum supremacy regarding encryption ("Q-Day") will likely occur between 2032 and 2033.

Interview with Anton Osicka, Founder of Lovable

• Reported Lovable has reached $600 million in revenue in 20 months, with over $500 million in revenue recorded by May of the current year.
• Highlighted that the platform sees over 1 million new projects built weekly and supports more than 50 million apps currently.
• Noted that 95% of Lovable's users are non-technical, though the platform is also heavily utilized by engineers for rapid prototyping and internal tooling.
• Described a shift from "vibe coding" for mockups to generating fully functional, secure, and production-ready software in days, reducing development costs from $500,000 (historical estimate) to under $2,000.
• Stated that Lovable has replaced an average of 10 existing enterprise tools (e.g., Salesforce, HubSpot, Slack) for some users, saving over $1 million annually in licensing and maintenance.
• Explained the company's "co-opetition" model, encouraging multiple internal teams to build competing versions of the same tool rapidly, then integrating the best features, a strategy enabled by low engineering bottlenecks.

Lovable's Technical Strategy and Business Model

• Confirmed a multi-model strategy that routes user requests to the most suitable frontier model (commercial or open-weight) based on task suitability.
• Revealed an investment in in-house "post-training" and reinforcement learning to fine-tune models specifically on mistakes made during customer usage, creating a proprietary intelligence loop.
• Reported that approximately 60% of customers hit token usage caps and purchase overages, driven by high addiction to the platform's speed and value.
• Announced the rollout of a hosting product line acting as an AWS competitor, allowing users to run software natively on the platform.
• Described the next evolution of the platform as an "AI Co-founder," capable of analyzing business data to provide strategic recommendations and optimizations even when users are offline.
• Clarified that while bespoke software is replacing some legacy tools (like internal SLACK or HR systems), Lovable maintains interoperability with major suites (Google, Microsoft) for data flow and security.
• Noted that the platform is profitable or close to it, prioritizing customer intelligence over using cheaper, lower-quality models.

Future Outlook for AI and Computing

• Predicted the convergence of classical computing, AI computing, and quantum computing will define the next era of technological acceleration.
• Argued that the next two decades will be the "best time in human history to be a technologist," with AI poised to solve major challenges in chemistry, biology, material science, and cancer research.
• Emphasized that while model capabilities (like Anthropic's "Fable" or Lovable's agents) continue to improve rapidly in generating code, the primary bottleneck remains human strategic decision-making and defining the "right" problems to solve.