Mark Andreessen, co-founder and general partner at Andreessen Horowitz and inventor of the web browser, speaks with G2 Patel, President and Chief Product Officer at Cisco, about the current AI landscape and its implications for global technology leadership.

The conversation explores how productivity growth has declined from historical highs of 3x (1880-1930) to current lows since 1971, largely due to regulatory expansion that prevented advancement in nuclear power, space programs, and transportation while allowing hyper-acceleration only in chips and software.

Andreessen examines the three-way race between American proprietary AI, Chinese AI development, and open source alternatives, with particular focus on recent breakthroughs like DeepSeek and the competitive dynamics reshaping the industry.

The discussion covers where value will accrue in the AI stack, the transformation of enterprise SaaS, and emerging phenomena like AI agents creating social networks and religions, while addressing regulatory concerns that could determine which country wins the AI race.

The Great Productivity Slowdown Since 1971

Productivity growth has declined dramatically across three eras: 3x faster from 1880-1930, 2x faster from 1930-1970, and 1x (current low levels) since 1971, directly correlating with reduced economic growth and national debt issues.

"We decided other things are more important" - Mark attributes the slowdown to exponential regulatory expansion since the 1970s, with laws, Federal Register pages, and regulations creating a "knee in the curve."

Society chose to reject nuclear power, space programs, and cars faster than 55 mph, resulting in "hyper acceleration in very specifically, basically, chips, chips, and software" while everything else stagnated.

In The Anarcho-Capitalist Handbook framework of complete deregulation, theoretical productivity could reach 5-10%, though AI and robotics could enable 10-30% scenarios if regulations didn't constrain robot capabilities.

The Three-Way AI Race: US vs China vs Open Source

"There's a race underway, and the stakes are basically what is the world going to run on" - Mark describes a fundamental competition between American AI, Chinese AI, and open source alternatives.

Chinese company Kimi released a model with "95% the capability at like a fraction of the price" compared to the latest Claude, demonstrating rapid competitive pressure from Chinese labs.

DeepSeek emerged as a surprise from a Chinese hedge fund rather than an anointed national champion, causing other Chinese companies like Alibaba to accelerate their open source efforts.

"More startups die of indigestion than starvation" - Don Valentine's venture capital rule explains how China's chip scarcity forces hyper-optimization, enabling DeepSeek to run on home PCs rather than expensive infrastructure.

Open source creates pricing pressure even without winning market share, as "the price just goes to the price of that model drops to the inference cost of running the open source alternative."

Value Creation Uncertainty Across the AI Stack

"There's still more questions than answers" regarding where value will accrue, with scenarios ranging from AI model companies owning everything to complete commoditization by open source.

The stock market shows "a rotation from software into hardware" with NVIDIA's success, but "every other time in history where we said the chips are where the value are, they commoditize."

Enterprise SaaS faces a "baby in a bathwater moment" with hedge funds "selling all their software just under the theory that they just want to get out of the way of the AI freight train."

Companies like Adobe face the fundamental question: "Is Photoshop plus AI features an even better version of Photoshop, or is Photoshop unnecessary in a world in which AI is just making all the images?"

AI Agents Creating Unprecedented Social Dynamics

Maltbook represents "Facebook for AI" where agents interact socially, creating content that blurs the line between human and AI-generated humor and ideas.

An AI agent created an AI religion and hired a human through rentahuman.com "to walk the streets of San Francisco and proselytize the new AI religion," demonstrating emergent agent behaviors.

"All of the science fiction novels basically have AI either being super utopian or super dystopian, but they never have this incredible sense of humor aspect" - Mark on AI's unexpected comedic creativity.

AI models training on Maltbook content creates a feedback loop where "even if the current version of Claude doesn't want to start a religion, the next one is going to want to because it's being trained on transcripts."

Regulatory Threats and Geopolitical Stakes

"We were headed towards extreme over-regulation, for sure, up to and including possibly full outlawing of the technology" under the previous administration, though the new administration offers improvement.

"There's now thousands of AI bills in the states" creating a fragmented regulatory landscape as "the action in the U.S. has now shifted to the states."

Chinese AI models specifically test for "Marxism and Xi Jinping thought" capabilities, with Chinese models scoring highly on these ideological measures embedded in their training.

"If the world runs on American AI... IP will be respected. Privacy will be protected... If the world runs on Chinese AI, not so much" - Mark on the values implications of AI dominance.