AI Venture Capitalist: These Tech Predictions Will Change Everything by 2030 w/ Vinod Khosla | #159
Peter Diamandis interviews venture capitalist Vinod Khosla on stage at the 2025 Abundance Summit about where AI, robotics, energy, and transportation are headed by 2030-2050. Khosla argues bipedal robots will dwarf the auto industry within 20 years, that AI will make expert-level knowledge free for doctors, teachers, and other professionals, and that programming itself is splitting into 'AI does it for you' (Cognition) and 'anyone can do it' (Replit) extremes. He makes the case that fusion (via Commonwealth Fusion) and super-hot geothermal will make energy abundant by 2030, and that the right fix for urban congestion is driverless public transit in dedicated lanes rather than more robotaxis. Khosla closes by rejecting resource scarcity outright, arguing untapped underground minerals mean abundance is a search problem, not a supply problem.
Bipedal robots vs. the auto industryRobotics▶ 3:01
Khosla says bipedal (or similar) robots will become a bigger business than global auto within 20 years, and the auto industry - except Elon Musk - hasn't noticed.
Every professional will effectively supervise about five AI interns within five years, with capability maturing over 3-10 years depending on the field.
Redesigning healthcare if expertise is freeHealth▶ 6:40
Khosla's question for healthcare boards: if all medical expertise were free, how would you design the system from scratch?
Pay equity when AI equalizes expertiseEconomy▶ 6:58
If a farm worker and an oncologist have the same AI-augmented expertise, Khosla asks how a national policy would decide how to pay people.
Advertising in an agent-mediated economyEconomy▶ 7:47
Diamandis and Khosla debate whether AI purchasing agents kill advertising; Khosla says it reverts ads to being genuinely informational rather than manipulative.
ChatGPT's paid tiers and OpenAI's Deep Research tool are shown via two anecdotes matching or beating expert human output: an 18-page biology report and a retail strategy review.
Future of programming: Cognition vs. ReplitAI▶ 15:07
Khosla frames programming's future as both 'no more programmers' (Cognition's AI that codes for you) and 'everyone's a programmer' (Replit's plain-English coding).
Computers learning humans, not vice versaAI▶ 17:03
Khosla says computing is flipping from humans adapting to software (SAP, Oracle, Workday) to software adapting to humans, which will make menu-driven UIs disappear.
Khosla describes three technologies he uses daily: Fountain Life full-body diagnostics, Viome gut microbiome testing, and OneSkin's senescent-cell peptide skincare.
Energy outlook: fusion and geothermalEnergy▶ 20:44
Khosla is bullish on energy: Commonwealth Fusion should hit net energy by early 2027 and existing coal/gas plants can be repowered with fusion boilers; super-hot geothermal drilling could beat natural gas on cost before 2030.
Fixing urban transportation with driverless public transitTransportation▶ 25:30
Khosla argues robotaxis worsen congestion like Uber did; the fix is driverless, on-demand public transit in dedicated bike-lane-width corridors that can carry 10x a street's current capacity.
Resource abundance and deep-earth miningOther▶ 28:56
Khosla says mineral scarcity is a search problem, not a supply problem - humanity has mostly mined surface deposits, and drilling a kilometer down would reveal abundant cobalt, lithium, and other metals.
Predictions made
openVinod Khosla: Bipedal robots (or similar form factors) will become a larger industry than the global auto industry within 20 years.
“By 2030 nobody will be debating whether fusion's a economic possibility”
Your call:
openVinod Khosla: The US will replace coal and natural gas power plants by retrofitting their boilers with fusion heat rather than building standalone new fusion plants.
“we will replace every coal and natural gas plant in this country by just repowering the boiler”
Your call:
openVinod Khosla: Super-hot geothermal (drilled far hotter than today's wells) will be deployed at scale, undercutting natural gas on cost, before 2030.
“possibly AI alone. Yeah. Uh I think that will happen.”
Your call:
Numbers that matter
Khosla invests across any AI-automatable job category with 100,000+ workers per US Bureau of Labor Statistics dataHis investment screening criterion for 'free expertise' categories like AI doctors, teachers, accountants.
Every professional gets roughly 5 AI interns within 5 yearsKhosla's framework for near-term AI leverage per professional, maturing over 3-10 years.
ChatGPT's $20/month tier already out-earns Google/Facebook per userKhosla citing consumer willingness to pay as proof AI subscription revenue can beat ad revenue per user.
OpenAI's premium tier runs up to $200/monthReferenced as the premium tier Khosla calls 'incredibly valuable.'
Roughly 40 VC-funded fusion companies exist todayUp from essentially zero investor interest when Khosla funded Commonwealth Fusion in 2018.
Drilling geothermal wells to 400-450F instead of 200-250F yields about 10x the power outputThe basis for 'super-hot geothermal' becoming cheaper than natural gas.
A bike-lane-width driverless transit corridor can carry 10x the capacity of a regular street lane, more than light railKhosla's pitch for his driverless public-transit investment as the fix for urban congestion.
The human body carries about 40 trillion human cells alongside about 100 trillion bacterial/fungal cellsDiamandis citing the scale of the microbiome that Viome's testing analyzes.
Worth digging into
🕳️ The 'Situational Awareness' paper
Khosla calls it extraordinary and tells the whole audience to read it (or feed it to NotebookLM) before discussing AI timelines further.
🕳️ The 'row 37' example as a metaphor for AI scientific leaps
Khosla cites it as proof AI can make creative leaps humanity never found by building incrementally on prior work.
🕳️ Super-hot geothermal drilling technology
Khosla says two separate efforts are tackling the metallurgy problem of drilling at 400-450F, and that it will 'surprise us.'
🕳️ Viome's metatranscriptomics origin story
Khosla says the underlying tech was developed in New Mexico - where the nuclear bomb was built - originally as a biodefense tool.
He calls it one of his favorite 2018 bets alongside OpenAI and Commonwealth Fusion, with a design that reportedly beats light rail in a bike-lane footprint.