Taiwan has less than 3 weeks of energy reserves — a blockade alone, with no shots fired, would cause an economic impact worse than the Great Depression, says former Intel CEO Pat Gelsinger.
All-In with Chamath, Jason, Sacks & Friedberg
Taiwan has less than 3 weeks of energy reserves — a blockade alone, with no shots fired, would cause an economic impact worse than the Great Depression, says former Intel CEO Pat Gelsinger.
TL;DR
Former Intel CEO Pat Gelsinger and Lovable founder Anton Osika join Jason Calacanis for two sharp conversations. Gelsinger traces Intel's decline to non-technical leadership, $100B in buybacks instead of R&D, and missing the foundry model — while warning that Taiwan's less-than-3-week energy reserves make a blockade more economically devastating than the Great Depression [1] — Pat Gelsinger "Taiwan has <3 weeks of energy reserves: Taiwan holds less than 3 weeks of energy reserves, meaning a Chinese blockade alone — without firin…" 16:03 . Osika reveals Lovable hit $500M ARR in May 2026, is generating a million new app projects per week, and is evolving from a builder tool into an AI co-founder for running entire businesses [2] — Anton Osika "Lovable hit $500M ARR in May 2026: Lovable reached $500 million in annualized revenue by May 2026, just 20 months after launch, growing by …" 40:09 .
Former Intel CEO Pat Gelsinger discusses Intel's decline, the geopolitical threat of a Taiwan blockade, and the AI buildout. Then Lovable founder Anton Osika shares the platform's explosive growth — $500M ARR, 1 million new apps per week — and the vision for Lovable as an AI co-founder for businesses.
The episode opens with Jason and Pat trading quick biographical notes — Gelsinger joined Intel at 18 and spent 34 years there, experiencing the company's golden era under legends like Andy Grove and Gordon Moore. Jason frames the conversation as a postmortem on one of America's greatest technology companies. Before diving in, the Airwallex sponsor read positions the platform as an AI-native alternative to legacy global payments infrastructure.
Pat Gelsinger delivers a frank autopsy of Intel's fall from dominance. The core diagnosis: when the company stopped being run by deeply technical leaders and handed control to finance executives, every decision ran through a spreadsheet rather than an engineering vision. By the time Gelsinger returned in 2021, Intel had gone almost a decade without building a new factory, had failed to acquire EUV machines, and had distributed $100 billion to shareholders — capital that could have maintained its manufacturing lead [1] — Pat Gelsinger "Intel gave $100B to shareholders: In the 5–6 years before Gelsinger returned as CEO, Intel returned $100 billion to shareholders via divide…" 05:11 . Gelsinger then describes the moment he realized Apple's departure was inevitable: when Intel offered to help Apple port macOS to x86, Steve Jobs calmly revealed he had already been quietly running that port for the past four OS releases, years before anyone at Intel suspected a chip switch was coming [2] — Pat Gelsinger "I've been working on that for the last 4 releases. He had been preparing the core technologies inside of Apple for something that might hap…" 06:38 . Jobs embodied the patient, parallel preparation that Intel's finance-led leadership had abandoned. The segment closes with Gelsinger reflecting on NVIDIA — a company Intel dismissed as making gaming toys — and his own killed Project Larrabee, Intel's GPU competitor that was shut down the week he left the company.
Jason challenges Gelsinger on whether the AI buildout is a bubble, pointing to eye-watering valuations and potential overcapacity. Gelsinger's answer is structural: nobody builds a data center without energy, and global energy capacity is only growing at 4–5% per year after a decade of US stagnation. That constraint is the natural ceiling on how hyped the market can get. Beyond the supply side, he argues the value of intelligence is nearly infinite — better logistics, finance, healthcare, and labor productivity all accrue from cheaper tokens — and cites Jevons' Paradox as evidence that lower token costs drive more usage, not less. He foresees periodic SaaS-style 'apocalypses' as industries get disrupted, but views these as healthy corrections rather than a bursting bubble. The conversation then shifts to quantum computing, where Gelsinger breaks from 25 years of perpetually-delayed predictions and sets a hard deadline: meaningful results across multiple industries before 2030 [1] — Pat Gelsinger "Quantum supremacy before 2030: Pat Gelsinger predicts meaningful quantum computing results will arrive before 2030 — approximately 40 month…" 22:48 . He cites proven qubit construction, error correction, and algorithms across multiple hardware modalities — trapped ions, photonic, spin — as evidence that engineering scale is the only remaining challenge. He flags encryption-breaking Q-Day implications arriving around 2032–2033.
Jason introduces Anton Osika with genuine enthusiasm, calling him one of his favorite founders, and Osika immediately justifies that billing with jaw-dropping numbers. In just 20 months since Lovable launched, users are creating one million new software projects every single week; the platform has hosted more than 50 million total apps; and those apps collectively receive over 700 million visits per month [1] — Anton Osika "1 million new apps per week on Lovable: Lovable users are creating one million new projects every single week on the platform, with over 50…" 26:34 . By May 2026, Lovable had reached $500 million in annualized revenue, growing by roughly $100 million every six months. Osika then turns to the customer profile: four out of five users have no technical background [2] — Anton Osika "80% of Lovable users are non-technical: Four out of five Lovable users have no technical or engineering background, showing the platform is…" 27:59 . They're first-time founders, enterprise employees building side hustles, and small business operators using Lovable to figure out what to build before committing resources. The 20% who are technically trained appreciate the platform's opinionated architecture — automatic security scanning, best-practice payment setups, built-in guardrails — freeing engineers from routine decisions. Osika notes that the no-code tools of 10 years ago promised this but couldn't deliver; LLMs made the actual software good enough to deploy.
With competitors and foundation model labs threatening to make Lovable obsolete every six months, Osika walks through the strategy that has kept the company growing instead. Lovable routes every request to whichever model — commercial frontier or its own fine-tuned open-weight — is most suitable for that task. A Stockholm research team applies reinforcement learning specifically to mistakes frontier models make inside Lovable's agent harness, using the enormous signal from a million weekly projects to improve continuously. Critically, the team has never made a decision to use a cheaper model when it measurably performs worse for customers — margin optimization never overrides product quality. Jason floats the question of whether Lovable is profitable; Osika is careful, noting they monitor margins closely but prioritize intelligence investment. About 60% of lowest-tier subscribers hit their caps and top up — a sign of deep product-market fit. The episode closes with a philosophical exchange about parallel experimentation: when Jason admits his team built two separate intranets — one for the US and one for Japan — Osika draws on his time at CERN, where isolated teams work on the same particle accelerator independently to avoid anchoring bias, only sharing results at publication. Now that building costs approach zero, running duplicate software experiments is not waste — it's the optimal way to avoid local minima and find the best product.
Chapter 1 · 00:00
The episode opens with Jason and Pat trading quick biographical notes — Gelsinger joined Intel at 18 and spent 34 years there, experiencing the company's golden era under legends like Andy Grove and Gordon Moore. Jason frames the conversation as a postmortem on one of America's greatest technology companies. Before diving in, the Airwallex sponsor read positions the platform as an AI-native alternative to legacy global payments infrastructure.
Pat Gelsinger joined Intel at age 18 and spent 34 years there, growing up under mentors Andy Grove, Gordon Moore, and Bob Noyce.
Intel's decline wasn't about bad luck — it was structural. When technical founders gave way to finance-driven business leaders, every major investment decision started running through a spreadsheet instead of an engineering vision. The result: no new factories for a decade, no EUV machines, and $100B returned to shareholders instead of compounded into the future.
Chapter 2 · 01:41
Pat Gelsinger delivers a frank autopsy of Intel's fall from dominance. The core diagnosis: when the company stopped being run by deeply technical leaders and handed control to finance executives, every decision ran through a spreadsheet rather than an engineering vision. By the time Gelsinger returned in 2021, Intel had gone almost a decade without building a new factory, had failed to acquire EUV machines, and had distributed $100 billion to shareholders — capital that could have maintained its manufacturing lead [1] — Pat Gelsinger "Intel gave $100B to shareholders: In the 5–6 years before Gelsinger returned as CEO, Intel returned $100 billion to shareholders via divide…" 05:11 . Gelsinger then describes the moment he realized Apple's departure was inevitable: when Intel offered to help Apple port macOS to x86, Steve Jobs calmly revealed he had already been quietly running that port for the past four OS releases, years before anyone at Intel suspected a chip switch was coming [2] — Pat Gelsinger "I've been working on that for the last 4 releases. He had been preparing the core technologies inside of Apple for something that might hap…" 06:38 . Jobs embodied the patient, parallel preparation that Intel's finance-led leadership had abandoned. The segment closes with Gelsinger reflecting on NVIDIA — a company Intel dismissed as making gaming toys — and his own killed Project Larrabee, Intel's GPU competitor that was shut down the week he left the company.
Claims made here
In the 5–6 years before Pat Gelsinger returned as CEO, Intel returned $100 billion to shareholders via dividends and stock buybacks.
When Pat Gelsinger returned to Intel in 2021, TSMC was producing 5 times the wafer volume of Intel.
The US share of leading-edge semiconductor manufacturing has grown from approximately 12% to approximately 18% since the CHIPS Act was passed.
In the 5–6 years before Gelsinger returned as CEO, Intel returned $100 billion to shareholders via dividends and buybacks instead of investing in new factories or technology.
When Intel offered to help Apple port macOS to x86, Jobs dropped a bombshell: 'I've been working on that for the last 4 releases.' Jobs had been quietly preparing Apple's core technology for a potential future switch for years before anyone at Intel suspected. This foresight — building optionality in secret — is what made Apple Silicon inevitable.
At Intel's CPU peak, the company literally laughed at NVIDIA's machines. GPUs were for gamers, not serious compute. But Jensen Huang kept building a deeper software stack — CUDA, SIMD, multi-threading — until Japanese HPC researchers realized these 'graphics cards' could tackle the world's hardest workloads. Intel even had its own answer, Project Larrabee, killed the week Gelsinger left.
TSMC's original vision — become the factory for the entire semiconductor industry — seemed so niche that Intel didn't take it seriously. Intel was vertically integrated and proprietary; TSMC standardized everything and welcomed any customer. Steady progress and a demanding Apple drove them from irrelevant to producing 5x Intel's wafer volume by 2021, now 7x.
When Gelsinger returned to Intel in 2021, TSMC was already producing 5 times the wafer volume of Intel, and now the ratio has grown to 7-to-1.
The CHIPS Act has helped grow the US share of leading-edge semiconductor manufacturing from 12% to approximately 18%, though 50% remains the longer-term goal.
Chapter 3 · 15:19
Jason challenges Gelsinger on whether the AI buildout is a bubble, pointing to eye-watering valuations and potential overcapacity. Gelsinger's answer is structural: nobody builds a data center without energy, and global energy capacity is only growing at 4–5% per year after a decade of US stagnation. That constraint is the natural ceiling on how hyped the market can get. Beyond the supply side, he argues the value of intelligence is nearly infinite — better logistics, finance, healthcare, and labor productivity all accrue from cheaper tokens — and cites Jevons' Paradox as evidence that lower token costs drive more usage, not less. He foresees periodic SaaS-style 'apocalypses' as industries get disrupted, but views these as healthy corrections rather than a bursting bubble. The conversation then shifts to quantum computing, where Gelsinger breaks from 25 years of perpetually-delayed predictions and sets a hard deadline: meaningful results across multiple industries before 2030 [1] — Pat Gelsinger "Quantum supremacy before 2030: Pat Gelsinger predicts meaningful quantum computing results will arrive before 2030 — approximately 40 month…" 22:48 . He cites proven qubit construction, error correction, and algorithms across multiple hardware modalities — trapped ions, photonic, spin — as evidence that engineering scale is the only remaining challenge. He flags encryption-breaking Q-Day implications arriving around 2032–2033.
Claims made here
Taiwan has less than 3 weeks of energy reserves, as reported in the Wall Street Journal approximately 2 weeks prior to this episode.
When a semiconductor fab is turned off, it takes 90 days to come back online.
A brownout of Taiwan's semiconductor fabrication capacity would cause greater economic damage than the Great Depression.
China has blockaded the Taiwan Straits approximately 7 times over the last 4 years.
Global energy capacity is expanding at 4–5% per year, while the US had roughly a decade of 1% annual energy grid growth.
Taiwan has less than three weeks of energy reserves. A Chinese blockade — no military engagement, just cutting off oil and LNG — would brown out the island. Fabs take 90 days to restart once shut down. The global economic impact would exceed the Great Depression. This isn't hypothetical: China has blockaded the Taiwan Straits 7 times in the last 4 years.
Taiwan holds less than 3 weeks of energy reserves, meaning a Chinese blockade alone — without firing a single shot — could brown out the island and halt its semiconductor fabs.
A brownout of Taiwan's semiconductor fabs would cause an economic impact greater than the Great Depression, according to Gelsinger, because fabs take 90 days to restart after being turned off.
China has blockaded the Taiwan Straits approximately 7 times over the last 4 years, demonstrating that a blockade scenario is not theoretical.
The AI buildout has a natural circuit breaker: energy. Nobody builds data centers without power, and global energy capacity is expanding at just 4–5% per year. That constraint limits how overheated the market can get. Meanwhile the value of a token — a unit of intelligence — is nearly infinite if it improves logistics, finance, and labor. Gelsinger sees two decades of growth, not years.
Quantum has been '5 years away' for 25 years — but Gelsinger says this time is different. We now know how to build qubits, error-correct them, and write algorithms against them. Multiple modalities (trapped ions, photonic, spin) are all showing results. Engineering scale is the only remaining challenge, and he expects quantum supremacy across multiple industries before 2030, with encryption-breaking Q-Day around 2032–2033.
Pat Gelsinger predicts meaningful quantum computing results will arrive before 2030 — approximately 40 months away — solving problems that are impossible to compute today.
Chapter 4 · 25:00
Jason introduces Anton Osika with genuine enthusiasm, calling him one of his favorite founders, and Osika immediately justifies that billing with jaw-dropping numbers. In just 20 months since Lovable launched, users are creating one million new software projects every single week; the platform has hosted more than 50 million total apps; and those apps collectively receive over 700 million visits per month [1] — Anton Osika "1 million new apps per week on Lovable: Lovable users are creating one million new projects every single week on the platform, with over 50…" 26:34 . By May 2026, Lovable had reached $500 million in annualized revenue, growing by roughly $100 million every six months. Osika then turns to the customer profile: four out of five users have no technical background [2] — Anton Osika "80% of Lovable users are non-technical: Four out of five Lovable users have no technical or engineering background, showing the platform is…" 27:59 . They're first-time founders, enterprise employees building side hustles, and small business operators using Lovable to figure out what to build before committing resources. The 20% who are technically trained appreciate the platform's opinionated architecture — automatic security scanning, best-practice payment setups, built-in guardrails — freeing engineers from routine decisions. Osika notes that the no-code tools of 10 years ago promised this but couldn't deliver; LLMs made the actual software good enough to deploy.
Claims made here
Lovable has been on the market for 20 months since its launch.
Lovable sees one million new projects created on its platform every single week.
Applications built on the Lovable platform collectively receive more than 700 million visits per month.
More than 50 million apps have been built on the Lovable platform since its launch.
Approximately 80% of Lovable users have no technical or engineering background.
Some Lovable users are building businesses generating more than $1 million in annual revenue on the platform.
One million new projects every week. Over 50 million total apps built. 700 million monthly visits to those apps. $500 million in annualized revenue by May 2026 — all in 20 months since launch. These aren't vanity metrics; they represent a platform where businesses are actually running. Lovable isn't a prototype tool anymore.
Lovable users are creating one million new projects every single week on the platform, with over 50 million apps built in total since launch.
Applications built on Lovable collectively receive more than 700 million visits per month, demonstrating significant real-world deployment and usage.
Four out of five Lovable users have zero engineering background. They're first-time founders, enterprise employees building side hustles, and small business owners figuring out what to build. The 20% who are technical use Lovable because it enforces best practices automatically — secure payments, security scans, architecture guardrails — things even developers appreciate not managing manually.
Four out of five Lovable users have no technical or engineering background, showing the platform is primarily empowering non-developers to build software.
Jason Calacanis's Founder University program needed an intranet. Two years ago that would have been a $500,000 project — off the table. Instead, a non-technical program manager built it herself on Lovable in 4 to 8 hours, including economic impact calculators, without asking permission. Total cost: under $2,000. The software now powers the program in Saudi Arabia and Japan.
Jason Calacanis's team built a full intranet for Founder University in 4–8 hours using Lovable for less than $2,000 — a project that would have cost $500,000 two years earlier.
Chapter 5 · 33:38
With competitors and foundation model labs threatening to make Lovable obsolete every six months, Osika walks through the strategy that has kept the company growing instead. Lovable routes every request to whichever model — commercial frontier or its own fine-tuned open-weight — is most suitable for that task. A Stockholm research team applies reinforcement learning specifically to mistakes frontier models make inside Lovable's agent harness, using the enormous signal from a million weekly projects to improve continuously. Critically, the team has never made a decision to use a cheaper model when it measurably performs worse for customers — margin optimization never overrides product quality. Jason floats the question of whether Lovable is profitable; Osika is careful, noting they monitor margins closely but prioritize intelligence investment. About 60% of lowest-tier subscribers hit their caps and top up — a sign of deep product-market fit. The episode closes with a philosophical exchange about parallel experimentation: when Jason admits his team built two separate intranets — one for the US and one for Japan — Osika draws on his time at CERN, where isolated teams work on the same particle accelerator independently to avoid anchoring bias, only sharing results at publication. Now that building costs approach zero, running duplicate software experiments is not waste — it's the optimal way to avoid local minima and find the best product.
Claims made here
Lovable reached $500 million in annualized revenue in May 2026.
Approximately 60% of Lovable's lowest-tier subscribers hit their usage caps and pay additional top-up fees.
Lovable is evolving beyond software creation into full business operations. In pre-release, users can access an AI co-founder that monitors their business overnight and delivers strategic recommendations each morning. With all your apps running on the platform, Lovable has access to all the data — usage, revenue, customers — to recommend optimizations without being asked. The product moat just got a lot deeper.
When a US nursing education company used Lovable to build custom scheduling, certification management, and admin tools, they didn't just save money — they replaced 10 separate software subscriptions and now save over $1 million per year. As building costs approach zero, every company will evaluate whether Salesforce, Slack, and HubSpot are still worth their price — or whether a bespoke alternative built in days beats them on fit and cost.
A company called Narsa replaced more than 10 internal tools with bespoke Lovable-built applications, saving over $1 million per year in software costs.
Lovable reached $500 million in annualized revenue by May 2026, just 20 months after launch, growing by roughly $100M every 6 months.
Lovable uses multiple frontier models and increasingly its own fine-tuned open-weight models, routing each request to whatever is most suitable. The team explicitly refuses to optimize for cost by using measurably worse models. A research team in Stockholm does post-training using reinforcement learning on the specific mistakes frontier models make inside Lovable's agent harness. The data flywheel from a million weekly projects is the core competitive moat.
Approximately 60% of Lovable's lowest-tier subscribers hit their usage caps and pay extra top-up fees, indicating high addictiveness and ROI for users.
When Jason's team built two separate intranets in Japan and the US without coordinating, he worried they'd diverged. Anton said: that's the right approach. At CERN, Anton witnessed isolated teams working on the same particle accelerator publish separately to avoid anchoring bias — free markets work by competition, not consensus. Now that building is cheap, running parallel experiments beats planning a single perfect product.
No indexed bits in this chapter.
Taiwan has less than three weeks of energy reserves. A Chinese blockade — no military engagement, just cutting off oil and LNG — would brown out the island. Fabs take 90 days to restart once shut down. The global economic impact would exceed the Great Depression. This isn't hypothetical: China has blockaded the Taiwan Straits 7 times in the last 4 years.
One million new projects every week. Over 50 million total apps built. 700 million monthly visits to those apps. $500 million in annualized revenue by May 2026 — all in 20 months since launch. These aren't vanity metrics; they represent a platform where businesses are actually running. Lovable isn't a prototype tool anymore.
This episode
Apple CEO described as a ruthless, visionary leader who secretly prepared Apple's OS for a chip transition years before revealing it to Intel.
US legislation providing semiconductor manufacturing subsidies, cited as having helped grow the US share of leading-edge chip production from 12% to 18%.
Legendary Intel CEO cited by Gelsinger as a deeply technical mentor who exemplified the engineering-led leadership Intel later abandoned.
NVIDIA CEO credited with building CUDA and transforming GPUs from gaming hardware into the backbone of AI computing.
Intel co-founder and originator of Moore's Law, cited as part of the deeply technical founding generation Gelsinger admired.
Central subject of the first half — its decline from semiconductor dominance is dissected by former CEO Pat Gelsinger.
Vibe coding platform founded by Anton Osika, discussed at length for its growth to $500M ARR, AI co-founder features, and potential to replace SaaS tools.
Discussed as the dominant global foundry that now produces 7x Intel's wafer volume and whose Taiwan location poses a critical geopolitical risk.
Discussed as Intel's demanding customer that secretly prepared its own silicon for years, ultimately launching Apple Silicon and cutting Intel out.
Held up as the company Intel dismissed as a gaming-GPU maker, which then dominated AI compute through CUDA and GPU architecture.
AI lab discussed as a partner and model provider for Lovable, maker of the Claude 'Fable' model used in the platform.
Particle physics research center where Anton Osika worked, used as an illustration of coopetition through parallel independent research teams.
Quantum computing company in Pat Gelsinger's investment portfolio, building photonic quantum computers.
Discussed as the geopolitically vulnerable home of TSMC, with less than 3 weeks of energy reserves making a Chinese blockade an existential economic threat.
Discussed as the geopolitical threat to Taiwan, with 7 blockades of the Taiwan Straits in 4 years raising fears of a future semiconductor supply chain crisis.
Stats
This episode
Factual claims made this episode, and whether a source was named.
In the 5–6 years before Pat Gelsinger returned as CEO, Intel returned $100 billion to shareholders via dividends and stock buybacks.
When Pat Gelsinger returned to Intel in 2021, TSMC was producing 5 times the wafer volume of Intel.
The US share of leading-edge semiconductor manufacturing has grown from approximately 12% to approximately 18% since the CHIPS Act was passed.
Taiwan has less than 3 weeks of energy reserves, as reported in the Wall Street Journal approximately 2 weeks prior to this episode.
When a semiconductor fab is turned off, it takes 90 days to come back online.
A brownout of Taiwan's semiconductor fabrication capacity would cause greater economic damage than the Great Depression.
China has blockaded the Taiwan Straits approximately 7 times over the last 4 years.
Global energy capacity is expanding at 4–5% per year, while the US had roughly a decade of 1% annual energy grid growth.
Lovable has been on the market for 20 months since its launch.
Lovable sees one million new projects created on its platform every single week.
Applications built on the Lovable platform collectively receive more than 700 million visits per month.
More than 50 million apps have been built on the Lovable platform since its launch.
Approximately 80% of Lovable users have no technical or engineering background.
Some Lovable users are building businesses generating more than $1 million in annual revenue on the platform.
Lovable reached $500 million in annualized revenue in May 2026.
Approximately 60% of Lovable's lowest-tier subscribers hit their usage caps and pay additional top-up fees.
We use essential and analytics cookies to run Vuci. To understand how the site is used: Privacy Policy.
Install Vuci on your phone
Add it to your home screen for a faster, app-like experience.
Install Vuci on your phone
Tap the Share button, then “Add to Home Screen”.
A new version is available
Reload to get the latest Vuci.