AI Venture Capitalist: These Tech Predictions Will Change Everything by 2030 w/ Vinod Khosla | #159
Summary
- Khosla’s major physical-economy call is that bipedal robots will exceed the auto industry within 20 years, with one billion by 2040 now “an underestimate.” At the $10,000 price point Diamandis cited, machines working 24/7 could perform more work than today’s manual labor across seven billion people; Diamandis said Elon expects Optimus to become Tesla’s largest business and said he believes that.
- AI makes expertise abundant and could reduce professionals’ roles. Over the next five years, Khosla urges every doctor, accountant or other professional to consider supervising roughly five AI interns; whether they mature in three, five or ten years will vary, but the strategic question is already: “If all expertise were free, how would you design a health care system?”
- The immediate AI opportunity is organizational deployment, because Khosla says people are using the technology at only “a fraction of your potential.” Deep Research separately returned an 18-page specialist report with the same two disease recommendations, prompting Khosla to ask whether AI is open during every board and executive meeting and receiving the same challenges as human teams.
- Consumer agents may transform advertising from emotional persuasion into decision-grade information rather than eliminate it. An AI can filter the pitch for $300 jeans when “$30 Levi’s will do just fine”; if the $20 ChatGPT subscription model holds, Khosla said it would imply more revenue per user than Google or Facebook, and he called the $200 tier “incredibly valuable.”
- Programming bifurcates into autonomous software workers and universal creation by natural language. Cognition represents the “software intern” that does programming, while Replit lets non-programmers build applications in English; Khosla’s larger inversion is that “instead of humans learning computers, computers will learn humans.”
- Fusion and superhot geothermal support Khosla’s optimism that energy constraints can be overcome. Commonwealth Fusion expects positive net energy output by early 2027, while 400–450-degree geothermal could yield 10 times the power of today’s 200–250-degree wells and become cheaper than natural gas before 2030.
- Autonomy is valuable when it expands transport capacity, not merely when it removes drivers. Khosla’s proposed driverless public-transit system could carry more than light rail in a bicycle-lane-width corridor and increase street capacity 10-fold; paired with exploration a kilometer underground for minerals, it embodies his thesis that technology converts apparent constraints into abundance.
Deep dive
1. Robotic labor scales beyond the auto industry
Diamandis pinned down Khosla’s earlier markers: one billion bipedal robots by 2040 and a later suggestion that a robot could cost $10,000. Khosla called one billion “an underestimate” and said the category, across form factors, “will be larger than the auto industry within 20 years.”
His arithmetic is utilization: robots work 24/7 rather than eight-hour shifts with breaks, so one billion could perform more work than all manual labor across today’s seven billion people. The auto industry has largely failed to recognize its biggest opportunity; Diamandis said Elon expects Optimus to become Tesla’s largest business and said he believes it.
2. Free expertise could reshape professional-service economics
Looking back to his 2012 essays “Do We Need Doctors?” and “Do We Need Teachers?”, Khosla said doctors will play a minor health-care role, including in certain forms of human connection, which he said will remain important. The same direction applies to teachers, oncologists, structural engineers and accountants.
His five-year operating model is that every professional should consider getting five AI interns, comparable to newly graduated junior staff. Their maturation might take three, five or ten years depending on the field; wherever U.S. labor statistics show more than 100,000 workers in a category, Khosla argues for investing in that category.
The end-state question is institutional, not technical: “If all expertise were free, how would you design a health care system?” At national-policy scale, if a farm worker and oncologist can access exactly the same AI-upgraded expertise, that raises the question of how people should be paid.
3. AI reshapes buying, research, coding and science
Khosla prefaced the forecasts with his governing hedge: “The only thing we know about the future is it’ll be uncertain. We have to figure it out in increments.” When Diamandis suggested consumer agents would ignore advertising, Khosla countered that ads may revert to information while AI filters attempts at “making you buy things you had no intention of buying.”
Khosla argued that consumer willingness to pay challenged old assumptions: if the $20 ChatGPT subscription model holds, it would imply more revenue per user than Google or Facebook, while the $200 tier was “incredibly valuable.” His conviction preceded that comparison—his 2018 OpenAI commitment was twice his previous largest initial investment.
Khosla’s chief of staff, a Ph.D. in viral immunology, produced an 18-page developmental-biology report recommending two disease indications. Deep Research returned a separate 18-page report with the same two disease indications. His executive challenge: keep AI open in every board meeting and give it “the same challenges you’re giving your team.”
Programming’s future is “both” no more programmers and everyone programming: Khosla cited his investments in Cognition, which supplies an intern that writes software, and Replit, which lets anyone create applications in English. Menus disappear as computers learn humans; Khosla said future Nobel Prizes could possibly come from AI alone, though whether humans would recognize AI as capable of winning one is a human limitation. The non-incremental creativity illustrated by the “row 37” example is a massive accelerator for humanity.
4. Fusion and superhot geothermal make energy abundance plausible
Khosla funded Commonwealth Fusion in 2018, when others dismissed fusion as perpetually 50 years away—but conceded, “I wasn’t that visionary” about connecting it to AI demand. The company expects positive net energy output by early 2027 and construction of its first plant soon afterward; it has announced a site for it. By 2030, Khosla expects economic feasibility to be settled, though cost questions may remain.
His deployment forecast is retrofit-led: replace coal and natural-gas plants by repowering their boilers with fusion, using existing locations and avoiding the 10-year permitting, grid-connection and power-line timelines he cited. “You just retrofit the same locations.”
Superhot geothermal offers a nearer parallel: raising well temperatures from roughly 200–250 degrees to 400–450 degrees can produce 10 times the power from the same well. High-temperature drilling and hard-rock fracturing remain difficult, but Khosla expects deployment before 2030 at costs below natural gas.
5. Autonomous transit multiplies the scarce resource: street space
Private autonomous cars repeat Uber’s congestion problem, only without drivers. Khosla instead proposes personal, driverless public-transit vehicles that show up when needed—even after a 1:00 a.m. restaurant shift—without requiring a schedule. They would travel to the destination without intermediate stops; off the fixed routes, a Waymo might serve the last mile.
Street width is, he argued, the one fixed constraint in cities; everything else can change. The vehicles could run in protected, bicycle-lane-width corridors without interacting with other traffic. Khosla’s modeling indicates more capacity than light rail and a 10-fold increase in street throughput, supporting his forecast that he suspects most cars in most cities will be replaced by 2050 while people can live farther away and retain a 30-minute commute.
Diamandis closed on resource abundance, and Khosla agreed: perceived mineral constraints arise because humanity mostly mines convenient surface deposits. Looking a kilometer below Earth’s surface with better tools could reveal abundant cobalt, lithium and other metals beyond what people happen to encounter near the surface.