Zynga Founder: Consumer Is Not Investible Right Now - Thats Why You Should Build It

Predictions and Expectations:

• Mark Pincus expects that the opportunity to offer a "new internet treasure" or reinvent generic services is "highly likely" due to AI and agents, even though the consumer sector is currently "arguably not investable."
• Pincus believes that "trillion-dollar consumer services" and companies are yet to be invented, noting that current home screens are "half empty" regarding new digital life stacks.
• Pincus predicts that the "Consumer Revolution" will actually occur in 2029, based on the Opus 4.5 moment happening in December and the need for a three-order-of-magnitude shift in capabilities.
• Pincus expects that the cost of inference will see a 10,000x increase, aligning with Jeff Dean's prediction, which will make unlimited AI usage affordable for consumers.
• Pincus fears that if 90% of enterprises have not received benefit from AI by 2026 or within three to four years, the lack of progress would be concerning and suggest a "skill issue" or fundamental problems.
• Pincus believes that companies are currently "spending a million dollars a year on tokens" without achieving commensurate output, describing this as "token maxing" where the cost does not yet match the value.
• Pincus expects that "new" product features often function like the "back of the box" for cereal—they get the trial but may not drive long-term retention.
• Pincus believes that consumers are not yet at the point where AI is "intelligence on tap" like water, but rather that this state is approaching as costs drop.

Timelines and Milestones:

• Pincus states the Consumer Revolution is positioned for 2029.
• Pincus identifies the current moment as the "Opus 4.5 and later moment," tracing the shift in AI utility back to December.
• Pincus notes that the internet boom took "almost a decade" to become widely appreciated, citing Amazon's consistent quarterly growth starting in the fall of 2002.
• Pincus mentions having started his first company, Freeloader, in 1995.
• Pincus references a timeframe of two years for the cost of intelligence to drop from $100,000 to $1,000 per unit.
• Pincus expects to open source his "Gemini live um voice plug-in" "shortly."
• Pincus anticipates that in two years, the cost of intelligence will drop by orders of magnitude, potentially from $100,000 down to $10,000 and $1,000.
• Pincus predicts that the "time machine moment" for consumer AI will arrive in a couple of years.

Technology and Product Direction:

• Pincus plans to focus on building consumer products that leverage "AI and agents" to reinvent services like cameras, Uber, and Airbnb.
• Pincus intends to create an "always on AI intelligence" product where the AI listens to conversations and acts as a "smart other person at the table."
• Pincus suggests that the future of software development involves LLMs writing the code rather than humans writing code that calls LLMs, a shift that could reduce code volume by 10 to 20 times.
• Pincus predicts a shift toward "squandering tokens" in R&D, where companies use frontier models freely to build superior products.
• Pincus believes that the next major innovation will involve "unlimited AI" inside consumer apps, making services that currently cost $1,000 a month free.
• Pincus hypothesizes that games will eventually be created "pixel by pixel by the AI" while the user watches, driven by the cost reduction in compute.
• Pincus identifies "zero latency" as a key differentiator for future AI response times that belongs in the "new" category of the Proven Better New framework.
• Pincus expects that the "prosumer" market, specifically for the 20 million developers, will be a key area for AI investment and consumer-style tactics.

Market and Industry Outlook:

• Pincus believes that 90% of enterprises investing in AI have not yet received any benefit, attributing this to "confounding effects" and a lack of product integration.
• Pincus observes that investors are currently 180 degrees off from first principles, favoring enterprise over consumer despite the latter's future potential.
• Pincus states that the consumer market is currently "arguably not investable" but predicts the opportunity is "never been greater."
• Pincus anticipates that the cost of intelligence will follow a log graph similar to the iPhone, eventually becoming affordable for everyone.
• Pincus notes that the "abyss" between products is a time for expanding "taste zones" before the next wave of mass market adoption.
• Pincus believes that the "new meta" will be visible when services are offered with "free unlimited AI" inside them.

Company Plans:

• Pincus plans to "keep building consumer products" despite current market conditions.
• Pincus indicates he is working backward from a future where compute is free to identify the "primitives" needed for future consumer apps.
• Pincus describes a "proven better new" framework where a company might start with a proven product, add a "better" feature (e.g., price, speed), and test a "new" feature (e.g., always-on listening) as a hypothesis that might not catch.

Financial Guidance:

• Pincus mentions that some peers are spending $1 million a year on tokens, with Peter Steinberger reportedly spending $1.1 million a month.
• Pincus predicts that as compute costs drop, the "work of a thousand people" can be done for a similar total cost to current high-end models, effectively increasing efficiency without necessarily increasing spend.
• Pincus estimates that the cost of intelligence will drop from $100,000 to $10,000 to $1,000 within two years.

Risks and Caveats:

• Pincus warns that 90% of enterprise AI investment is currently failing to deliver benefit, and he might get "a little worried" if this persists beyond three or four years.
• Pincus cautions that "new" product ideas often fail to catch, and founders should be prepared for "none" of their novel features to succeed.
• Pincus notes the risk of "token maxing" where companies spend heavily on compute but do not achieve proportional output, resulting in a "skill issue."
• Pincus fears that the current "cost problem" with AI makes magical experiences unaffordable for the mass market, limiting them to enterprise-level spending of $1,000 a month.
• Pincus advises founders to prepare for the possibility that they are "wrong" about their specific product variant and to stay dispassionate about their ideas.
• Pincus observes that the game industry faces a significant cost barrier where "every cent counts" in freemium models, making AI-generated content currently "very expensive."

Confidence and Disagreement:

• Pincus is convinced that the opportunity to reinvent services with AI is "highly likely," stating "it's intelligence on tap" is coming.
• Pincus disagrees with the industry trend of pivoting consumer products to enterprise simply because they are "more fundable right now."
• Pincus expresses skepticism regarding the current state of Siri and Amazon's Alexa teams, noting it is "unbelievable" that they have not delivered real LLM experiences despite large teams.
• Pincus questions whether current enterprise AI adoption is a "skill issue" rather than a technology limitation, especially if the 90% failure rate persists.
• Pincus suggests that the "fish are running" (success signal) only happens a "couple of times" in a founder's life, implying it is hard to predict and often subjective.
• Pincus believes that "founder mode" is necessary for every founder, countering the view that it is only for a "very small percentage" of exceptional leaders.