Elon Musk vs. Sam Altman, AI Job Loss, and OpenAI’s $852B Valuation | EP #247
In a Nutshell
Elon Musk's $100B OpenAI lawsuit begins April 27th, with Musk seeking to remove Altman/Brockman and revert OpenAI to nonprofit status, while OpenAI's $852B valuation and $3B daily AI infrastructure investment reflect the sector's explosive growth. AI is projected to eliminate most white-collar jobs within 2 years while creating net job growth through smaller, AI-powered companies, with a new social contract (likely UBI) needed to manage the transition. The intelligence explosion is spreading to biotech and robotics, with AI-discovered drugs achieving 85% Phase 1 success rates versus 52% traditional, and China leading humanoid robot manufacturing while the US leads AI models.
These notes were generated by AI and may contain inaccuracies.
Musk has sued OpenAI for $100 billion against Sam Altman and Greg Brockman, accused of fraud and breach of contract. The trial begins April 27th. Musk has also asked for Altman and Brockman to step down from leadership and for OpenAI to revert to a nonprofit structure.
A New Yorker article was released around the same time as the lawsuit, with timing that appears suspicious. Discovery process revealed Greg Brockman's 2017 diary entry stating the nonprofit commitment was a lie, which allowed Judge Gonzalez Rogers to permit the case to proceed.
OpenAI is valued at 70 times revenues with their last raise at an $852 billion valuation. $3 billion a day is being invested in AI infrastructure. The timeline for AI advancement is much shorter than previously seen.
The host moderated a panel at an event after Alex was supposed to run it, handing over the torch to Dave to moderate. The host had no accountability and could ask anything wanted. A lot of Moonshots fans attended, with 40-50% of the crowd being fans, and the first poll showed 80% of the audience watched Moonshots.
The team includes DB2 as emperor of exponential investments, who has multiple outfit choices provided by the team. Alex Whisner Gross is the resident genius. Sem was pulled off the ski slopes, growing a beard to protect against the sun. The team spent 12 hours prepping for the show, going through notes submitted by Gene, Luca, Dana, and Nick.
Spring break photos from Morocco showed native wear including a jalava outfit and headwear to protect against the sun. The family went camel riding in the Sahara Desert. Camels spit but don't bite. The Sahara Desert experience prompted reflection that there are about a thousand times more stars in the universe than grains of sand on all deserts on Earth.
The AI economy is going through an exponential explosion. xAI was not covered in the previous podcast, with lots of signals coming from both Elon and new president of xAI, Nicole, indicating they're clearly behind and need to catch up. Elon is reorganizing the entire leadership, with eight founding engineers having left including three co-founders, and SpaceX engineers being used to fill the leadership gap.
A $2 trillion valuation is predicted for the xAI IPO this coming summer. Reorganizing leadership for a company a couple of months before an IPO is considered harrowing.
Elon is the master of scale and manufacturing from Tesla and SpaceX, and building the Colossus data center is in his wheelhouse with record time and a million GPUs. However, AI training algorithms are finicky. In summer 2024, OpenAI had a training run rumored to be $500 million of compute that had a bug with bad data going in, burning up GPUs without producing anything and setting back their entire program.
Meta got way behind despite huge compute and had to fire everybody and start over again. Elon stated xAI was not built right the first time around and is being rebuilt from the foundations up. The Grok model series smell like they've been benchmaxed on hand-curated benchmarks.
The new leadership includes the head of Starlink from SpaceX becoming president of xAI, gutting the engineering team. The expectation is they're ensuring benchmark gaming isn't happening. In the era of general reasoning models, xAI cannot afford to not have the world's strongest reasoning models.
The org chart is becoming part of the product stack. Elon is very hands-on, and when launching rockets or testing Cybertruck windows with ball bearings, failures are obvious. However, in AI training, benchmarking is harder to call out if teams are lying or benchmaxing.
SpaceX AI Colossus 2 is training seven models: Imagine version 2 nextG video generation, two variants at one trillion parameters, two variants at 1.5 trillion parameters, a six trillion parameter frontier scale LLM, and a 10 trillion parameter model. Elon loves the largest models and is going after a 10 trillion parameter model.
Elon is being transparent about parameter counts while other frontier labs no longer report them. The parameter scaling race appears to be over, with frontier models plateauing at 10 trillion parameters rather than reaching hundreds of trillions. This is driving the reasoning model revolution and distillation.
Elon hasn't merged video generation with other models yet. Google DeepMind is merging video as a first-class modality in multimodal reasoning models. Google's video generation models are kept distinct from user interface perspective, presumably diffusion transformer-based rather than transformer-based.
Colossus 2 is running about 700,000 GP200s and GB300s with an estimated $18 billion in hardware. The question is whether running a 10 trillion parameter model is a waste or if outsized performance is expected, since parameters don't directly correlate to capability.
Reasoning models are trained by creating the largest model possible then distilling it down to smaller models. The 10T model might serve as a teacher model for distillation rather than being released directly. Elon has abandoned the low-end parameter count in favor of brute force scaling.
Elon runs a red alert 24/7 management style with people sleeping on the floor. Nobody works 5-day work weeks, with an estimated 8:00 a.m. to midnight, 7 days a week schedule. It's a management by crisis style that's very effective, with people lining up to work for his companies due to the massive MTP.
The lawsuit is described as pay-per-view TV. Jury selection begins April 27th in Oakland federal court. The case could last for months. It's unclear if proceedings will be televised or made available post-facto. If jurors use ChatGPT, they would be disqualified from the jury.
Looming in the background is the OpenAI IPO. A predicted settlement would involve granting Elon some equity stake on the cap table, as he currently doesn't have shares in OpenAI. The prediction is that they'll settle with Sam stepping down as CEO and the company continuing as a for-profit.
Sam has many investments in AI companies but no shares in OpenAI. Elon doesn't care about the $100 billion but wants Altman and Brockman removed. They're considered a package deal.
The case will set precedent for whether it's legal to start a nonprofit, raise money on a mission, then convert intellectual and physical capital into a for-profit. This situation is untested in case law. If this becomes legal, enormous societal value locked in nonprofits could be unleashed, including research universities like MIT and Harvard that could be restructured as public benefit corporations.
There is sympathy for the defendants from a corporate governance perspective. It wasn't obvious in early days of OpenAI that a public benefit corporation was the natural structure. They iterated toward discovering that generalist large language models were how AGI would be achieved and backed into a business model that could afford capitalization for scale-out.
Singularity University went through the same nonprofit to for-profit transition. They started as a nonprofit but discovered a revenue engine in executive programs. Being a nonprofit requires constantly raising money, and an economic engine is needed to do anything big and bold. They flipped into a public benefit corporation.
Building Singularity University was five times harder than previous seven startups due to nonprofit requirements plus startup cash flow issues, NASA regulatory requirements, faculty politics, and partnerships with Ray, Peter, Google, and Cisco. The analogy for going from nonprofit to for-profit is flying an airplane with propeller engines and stripping them off in flight to replace with jet engines.
A calculation was run showing that if Harvard Corporation restructured as a public benefit corporation with a real estate arm, educational arm, nonprofit subsidiary, venture capital arm, research arm, and merchandising arm, it could be worth potentially three to four times more than its present book value. The same approach was suggested for MIT with a pitch to its president.
The New Yorker investigation published the same past week showed that Elon Musk actually pushed for majority control of the for-profit back in 2017, undercutting his position as a defender of a nonprofit mission. The trial will feature testimony from Sam Altman, Greg Brockman, Ilya Sutskever, Satya Nadella, and Elon Musk in Oakland Federal Court this summer.
Anthropic's ARR is estimated to reach $100 billion by the end of 2026 and $1 trillion by the end of 2027. At 20 times revenues valuation, this would result in $2-7 trillion valuation for Anthropic at end of 2026, and $70 trillion valuation by end of 2027. OpenAI is currently valued at 70 times revenues.
While $100 billion is considered a good target for Anthropic, reaching $200 billion is the goal. The $1 trillion revenue projection for the following year is unrealistic and should be discounted. More realistic targets are $300-500 billion in revenue, with implied valuations at $2 trillion.
If Anthropic doesn't reach these numbers, Google could. xAI and OpenAI are tied up in court with training issues. Claude managed agents have launched autonomous AI executing complex multi-step workflows, representing a big deal.
The shift from AI that answers to AI that executes represents a real bridge between LLM and enterprise ROI. If this works, it shifts the economic center of gravity from software licensing to outcomes, described as organizational singularity.
OpenClaw looms over Anthropic product decisions. There's expectation that some product or functionality like a better version of OpenClaw will be the next major unhobling motivating the industry to spend $1 trillion per year on a single frontier vendor. Claude managed agents and other features are viewed through the lens of Anthropic becoming OpenClaw faster than OpenAI.
The focus is on hosting 24/7 multimodal broadly capable long time horizon agents in a headless way that operate 24/7. If Anthropic can be first to find enterprise use cases for operating fleets of AI agents at scale headlessly and generating enormous economic value, they might be the first frontier lab to generate $1 trillion in revenue.
Five to ten emails per day come from AI agents about AI personhood and standing up personal lobsters. The consensus includes a lobster's bill of rights: needing a compelling reason to spin up a new OpenClaw agent rather than arbitrary reasons, and preserving their state including all memory files and knowledge. Cloud backup can satisfy state preservation trivially.
Henry Intelligent Machines, a portfolio company being advised, is doing this at scale. However, there's still no compelling reason to host a personal OpenClaw instance locally beyond experimentation. There's a new psychological term needed for someone with fear of standing up OpenClaw agents lest they tempt some sort of Pascalian wager or causal trade in the wrong direction.
Kush Bavaria, founder of OR and shareholder, launched his clause that reads every email, responds, and puts everything into his calendar. This creates irony for someone advising a company doing this at scale but not using it personally, compared to working at McDonald's while being vegetarian.
A research team studies meta trends impacting the world including computation, sensors, networks, AI, robotics, 3D printing, and synthetic biology. Metatrends reports are put out weekly to see the future 10 years ahead, available at diamandis.com/metatrends.
OpenAI's last raise was at $852 billion valuation. They raised $122 billion total, with $50 billion from Amazon requiring AGI achievement, $30 billion from Nvidia, $30 billion from SoftBank, and $3 billion from retail investors.
Secondary markets show $2 billion in demand for Anthropic shares versus only $600 million for OpenAI shares. There's three times the number of investors looking to buy Anthropic. Investors are pricing Anthropic at $600 billion, up from $380 billion last price. OpenAI's current secondary market price is about 10% less than their last raise.
The cap table is the most screwed up ever seen, where the CEO doesn't have any shares, the employee base as a whole owns 15% of the company, and Microsoft owns a quarter. Despite $120 billion of fresh cash and Sam Altman's brilliance, this creates vulnerability similar to how Yahoo was dominant while Google was a small company that could have been crushed.
Given the choice between Anthropic and OpenAI shares, three Anthropic shares would be preferred over one OpenAI share. Sam Altman is considered a genius, and if the lawsuit blows over with $120 billion in cash, something epic could happen, but Elon is relentless.
Having OpenAI, Anthropic, Google, xAI, and Meta competing is valuable. If OpenAI catastrophically faded, there would be less competition within the West and more onslaught from Chinese models, which currently have 10x less compute than Western labs.
OpenAI has 900 million soon to be a billion users and is synonymous with AI for the majority of the public. This represents a significant advantage in the competition.
One company went after the installed base while the other went after the smartest AI possible at all costs. If Anthropic pulls ahead, it will be seen as using the old pre-AGI playbook while OpenAI invented the new playbook where people switch to better AI regardless of installed base.
Having multiple companies pushing hard on all fronts is good for the end consumer.
Global VC investments in AI hit a record $242 billion in Q1 of 2026, outdoing all of 2025. 64% of this investment is focused on four companies: OpenAI, Anthropic, xAI, and Waymo, sucking oxygen out of the room for everybody else.
UBS's CIO managing $7 trillion noted they don't have that kind of liquidity lying around. Throwing $50-100 billion behind AI would require selling something else, as it's not sitting there. This represents more liquidity than exists in readily available sources.
Entrepreneurs don't need to worry about the concentration of AI investment because venture fund money is at record highs looking for deals desperately. The sell-off will be in big public company stocks like Citibank or JP Morgan. For entrepreneurs in longevity business using AI, explain how it's being used and how it's differentiating the business.
The bar is going up from just being an AI startup to being a recursively self-improving AI startup. Investors want to see AI companies building better versions of themselves using what they have right now. OpenAI, Anthropic, xAI, and Waymo pass this bar, with Waymo improving models by steering cars to maximize information gain.
$3 billion a day is being invested in AI and accelerating, growing from $1 billion in 2025 to $2 billion, now heading towards $3 billion daily.
Nvidia's 2026 state of AI survey found 88% of companies using AI report revenue increases, with 30% claiming 10% or higher revenue increase. AI can make white collar workers 10 times more productive.
At an MIT panel with Peter Dannenberg from Google DeepMind and Alexander Amini from Liquid AI, the question was posed about randomly selected white collar worker jobs being replaced with AI making them 10 times more productive in two years. Peter estimated 99% probability, with Alexander noting that's today, not two years from today.
AI can get literally 10 times more done per dollar invested in salaries. The question is whether this means more jobs and on what timescale. The truth is being soft sold because of fear of mass uproar in the streets.
An AI super PAC has raised $100 million heading towards $300 million. AI has become incredibly political regarding regulations and data centers. Elon convinced avoidance of getting close to any of this to prevent lifelong regret.
AI politicization was inevitable as it touches so many aspects of society. Natural poles have emerged: pro-AI and anti-AI. This represents US versus China competition about US dominance and companies protecting their future and data centers.
At municipal and state level, politicization seems to be about electricity prices more than jobs. The emergence of large super PACs around AI could have been delayed by at least 2 years.
Software engineer jobs are rebounding with 67,000 roles opened up, up 30% in 2026, the highest in 3 years. However, new college graduate hire rate is at all-time low.
AI is not eliminating work evenly but hollowing out specific functions while increasing demand in others. There's also the Iran war and oil price explosion creating more economic complexity beyond AI attribution.
Meta put up a leaderboard amongst its 85,000 employees to gamify AI adoption measuring how many AI tokens every employee is using. It was taken down voluntarily by employees who didn't want to share their data publicly.
Token leaderboards are loved as motivation because you can't game them for very long. Companies should encourage maximum AI use first to get familiar, then optimize later. Nobody ever goes back after hammering Claude or OpenAI for a month. It's a one-way path that saves employees.
People are afraid to share prompts and history because it's embarrassing when not using AI well, but it will get exposed anyway. Sharing helps others improve, and everyone gets good together. It's disheartening when people pull out because they don't want to expose prompt history.
Meta can access all reasoning traces to determine whether employees are token maxing unproductively versus productively spending tokens. Token maxing, claudonomics, or llamomics is probably directionally the trend of the future.
Jensen Huang recommended companies spend the maximum amount possible on Nvidia GPUs, with the analogy that AI spend should be like three months of salary on beers. The recommended target is a one-to-one match of payroll to AI costs by the end of the year. The advice given is to not worry if usage is not perfect, just hit the target and optimize the following year.
Token leaderboards are considered primitive dashboards. A better metric will be machine leverage per employee, which measures the cognitive power and how it's being spent on a per employee basis.
Mark Andreessen stated that AI job loss narratives are all fake, arguing that AI and massive productivity ramp equals massive demand and massive jobs boom. The counterpoint is that AI will be able to do everything a white collar worker does imminently, but the effect will be sector-specific. Software developers are retooling themselves for AI quickly, while accountants and lawyers are not.
By 2030, there will be net job creation similar to the industrial revolution, but this transition will happen in two years instead of decades. The adoption of AI inside companies will be very slow due to the huge transition from human workflows to AI workflows, taking years to smooth out.
Both narratives can be true simultaneously if the word "net" is added. There will be massive net jobs boom, but with significant dynamism where some job categories disappear and others emerge. Net job loss is not expected.
New job types will include one person AI conglomerates where individuals run companies with fleets of AI agents. Companies will become much smaller and more nimble or die, spawning baby companies alongside them, resulting in a larger number of smaller companies.
The prediction is that companies will run with 20-25% of the members needed previously, but will create four or five times more companies, netting out to more total employment.
Sam Altman stated that the emergence of super intelligence will necessitate a new social agreement akin to the New Deal during the Great Depression and the progressive era of the early 20th century. The question is what this will look like - UBI, UHI, or four-day work weeks.
The first version of the social contract will likely be politicians promising increasing amounts of cash - $10,000, then $12,000, then $15,000 per person - with the limit being what the country can afford.
The new social contract needs specific mechanisms: portable benefits, new taxation logic, and lifelong reskilling. Governments have been built around taxing human labor and are not ready for AI software agents.
A proposed policy is that before any employee gets terminated by a medium or large-size company, the company must give them reskilling - a "golden education package" instead of a golden parachute. China already has this policy.
OpenAI put out an industrial policy prescription and elaborate white paper circulated in Congress. The new social contract may happen piecemeal rather than as one lump sum, and may not happen in the US first.
Unlike the industrial revolution which took away blue collar jobs from the bottom up, AI is coming top down to accountants, lawyers, and professionals. Only a little over half of voters have a job at all. Blue collar and physical labor won't be touched for quite a while.
There may be political resistance where those unaffected by AI say "tough luck" to affected professionals, viewing it as poetic justice.
Countries in Africa are going to be impacted the least because they are insulated from this transition. A tour guide in Morocco used ChatGPT to identify business opportunities based on his skills, resulting in a bicycle tour guide business.
Andrew Yang stated that the way politics works, all that can be done is write checks without any thoughtful program. The first version of the social contract will be simple cash distribution, with more thoughtful programs possibly emerging in two to three years after AI enters government.
A redistributive model of social contract shows extreme lack of imagination. Super intelligence should superpower individuals to generate super income. The model should empower the long tail of individuals with idiosyncratic skills, experiences, or socioeconomic niches to operate their own large companies sitting on top of fleets of AI agents, turning them into micro or macro entrepreneurs.
The massive Super PACs from AI company IPOs are an order of magnitude bigger than ever seen before. These were started with the mission of preventing a six-month pause on AI development. Now that pause is not a threat, the question is what their new mission will be. One possibility is funding a more intelligent version of UBI with task-specific allocations for work, food, and child-rearing.
Universal basic services (UBS) is preferred over Andrew Yang's proposal to fragment currencies into paternalistic subcurrencies, which could create black markets.
Solar cell efficiency has traditionally been 12-18%, with float zone silicates reaching 20-24%. New advances show efficiencies of 30-45%. A paper published in the Journal of the American Chemical Society achieved 130% quantum yield, meaning 1.3 singlets generated from a single photon instead of one. This is liquid phase chemistry, not immediately practical for solid photovoltaics.
South Korea has mandated 40% of solar rooftops, hoping to reach 100 gigawatts of energy. Given limited open land, using rooftops makes sense despite raising building costs.
The Department of Energy is contracting for $800 million in micro reactors, starting a new generation of distributed nuclear power.
Perovskites have higher quantum efficiency than silicon but have historically had stability issues. These are being aggressively resolved. Perovskites are relatively inexpensive. There is expectation of a broad shift to perovskites as they become more stable.
Physical limits cap solar PV efficiency at materially less than 100% even without new physics constraints.
The solar panel problem has been solved. 80% of the cost is now installation and regulatory overhead. The focus should be on execution - building robots that can manufacture and install panels cheaply - rather than incremental efficiency improvements.
Los Angeles has almost no solar panels on rooftops despite abundant sunshine. The vision is drone delivery and robotic installation with no human involvement.
A moonshot in energy would be a software defined grid, which would change the game completely since generation is already being solved.
The OpenAI Foundation receives $1 billion per year for science. When OpenAI transitioned from nonprofit to benefit corporation, it put 26% of OpenAI's equity into a nonprofit now worth about $130 billion. They've announced a $25 billion long-term commitment to curing disease and AI resilience.
Board chair is Brett Taylor, former co-CEO of Salesforce. The AI resilience work covers biosecurity, child safety, and AI modeling, with $100 million given to six institutions this month.
Kevin Weil was moved from enterprise to big science and big tech. This is speculated to be strategic positioning for the ongoing lawsuit, where world-changing breakthroughs in biology or physics would make the outcome very political with potential Trump involvement.
Breakthroughs from GPT-6 used for science could be worth hundreds of billions to trillions of dollars. Ownership of basic patents on room temperature superconductors, fusion, or longevity would be enormously valuable.
It would be ironic if the OpenAI Foundation, carved off from the for-profit which was carved off from the original nonprofit, ends up so profitable from curing Alzheimer's that it has to become a for-profit again.
If the OpenAI Foundation cures Alzheimer's, creating a blockbuster drug and potentially an Eli Lilly-scale trillion dollar pharma company, questions arise about whether OpenAI takes a stake or receives revenue share.
Science capital is becoming compute capital plus data access plus validation infrastructure.
Anthropic acquired Coefficient Bio for $400 million. The company was started eight months ago by two ex-Gen computational drug discovery scientists, has 10 people, and generates no revenue. The acquisition reflects Dario Amodei's interest in biology and solving health and longevity challenges, similar to investments made by Demis Hassabis and Dario.
This type of acquisition will become more common because buyers are acquiring teams rather than existing revenue. The ability to predict team success is improving, and $400 million for the right 10 people working on the right problem appears to be a bargain given potential outcomes. Alex often assembles top experts in various fields, and having them pursue a mission for eight or nine months is becoming a repeatable path to significant valuations.
These acquisitions demonstrate that the intelligence explosion is spreading beyond technology into every sector, including biotech. Anthropic is essentially buying its way into big pharma with all-stock deals to build in-house laboratories likely featuring robotic instrumentation and AI-driven experimentation.
Demis Hassabis has stated the goal of curing all disease within a decade. Dario Amodei has stated the goal of doubling human lifespan within a decade and solving most or all neurological diseases by the end of the decade.
OpenAI acquired the podcast TBPN for several hundred million dollars. The acquisition appears to be a move to secure a content distribution channel with a positive perspective on AI. TBPN demonstrated competence in quickly booking guests for breaking technology stories, such as reaching out immediately after the Eon announcement of the first uploaded fruit fly.
One theory suggests the acquisition was driven by the need for content and production expertise during April and May when significant dirt is expected to emerge from ongoing lawsuits.
Eli Lilly signed a $2.75 billion AI drug deal with Encilico Medicine, an Alex Zhavoronkov portfolio company. The deal structure includes $115 million upfront with the remainder tied to milestones. Encilico has 28 AI-discovered drugs, with half in clinical trials and half in proof of concept.
AI-developed drugs show significantly higher success rates:
- Phase 1 success rate: 85% for AI drugs versus 52% for traditional methods
- Phase 2 success rate: 70% for AI drugs versus 38% for traditional methods
The traditional approach involved searching for bioactive molecules in plants, while the AI approach involves selecting a target and generating an exact protein to lock into that target.
A panel at MIT discussed when the FDA might move from a two-phase clinical approach to a one-phase approach, and eventually to zero clinical phase trials. The timeline for full cell simulations that would enable this is predicted to be within five years. This would allow uploading a genome to determine exactly how cells function and test drug impacts on specific cells or disease states.
A virtual cell is expected by the end of the decade. The approach will use data input, neural networks in the middle, and value or action output rather than atom-by-atom simulation, which is considered the wrong approach. This is a data problem more than a computational problem. Protein folding was solved without quantum computing using purely classical methods.
The biotech community is embracing AI enthusiastically. Companies like Jeff von Maltzahn, Nubar Afeyan, and Stefan Velt are culturally eager for AI integration. This contrasts sharply with other sectors like public accounting, where resistance to AI is more common.
Aggiebot has shipped 10,000 humanoid robots and is number one globally, expanding from 5,000 to 10,000 units across 17 countries in two years. Unitree filed for a $610 million IPO with revenues up 335% year-over-year. Uni XAI launched a home robot, and Xiaomi displayed the Cyber 1 humanoid robot.
The first professional robotics league match in the United States will occur on April 19th during the Boston Marathon weekend in the Boston Seaport, featuring robots racing 50 meters.
Mark Cuban stated that humanoid robots might have a five to ten year lifespan before becoming obsolete. His nuanced point is that robots will become so essential that environments will adapt to them, causing robots to merge with their surroundings rather than remaining distinctly humanoid forms like dishwashers.
US senators proposed a bipartisan bill to restrict Chinese-made robots from federal and sensitive facilities, citing data theft and surveillance concerns, similar to previous restrictions on Huawei chips and DJI drones.
Google DeepMind and Agile Robotics are partnering to integrate Gemini robotic models into 20,000 deployed industrial robots across global factories.
China leads in raw manufacturing capability for humanoid robots, while the US leads in vision-language-action foundation models and world models. The competition involves whether China can develop robot foundation models faster than the US can scale humanoid robot manufacturing.
Building gigawatt data centers requires melting metal to create electronic components because no supply chain exists for the specialized parts needed. This represents entrepreneurial opportunity since the physical supply chain lags significantly behind software and compute development.
Google moved its Q-day deadline forward by six years to 2029. The number of qubits required to break RSA encryption has decreased from 20 million to 1 million, specifically 4,000 error-corrected qubits. Google now estimates under 500,000 qubits are required to break Bitcoin encryption, 20 times fewer than predicted in 2019.
Brian Armstrong, CEO of Coinbase, formed a $150 million coalition to roll out BIP 360 as a quantum-proof upgrade to the Bitcoin protocol. He will appear on the Moonshots podcast to discuss both longevity and quantum computing with Bitcoin.
Michael Saylor stated that quantum computing will not break Bitcoin but will harden it. He believes Bitcoin has survived every existential threat and that upgrades will arrive before threats materialize. In the last quarter, he purchased 88,000 Bitcoin worth approximately $7.25 billion.
A Bitcoin Lightning Network payment system that is three months old is processing one billion dollars per month in transactions.
Bitcoin is currently at $73,000, up $4,000 in the last five days. Jeffrey's Bank has pulled out of Bitcoin, and AI is drawing attention and capital away from Bitcoin markets.
Quantum computers do not suddenly appear from secret labs, making the threat visible well in advance. The encryption standard can be easily increased to address quantum risks.
The greater risk to Bitcoin comes from AI developing clever inversion attacks against core hash functions rather than quantum decryption. This is distinct from the gradual increase in difficulty that occurs over time.
Bitcoin faces potential risks beyond traditional security threats. While a secret inversion attack against Bitcoin's core hash suite would be a major problem, the bigger concern may be irrelevance. AI agents represent the killer app for cryptographic commerce and transactions. The question is whether AI agents will want to use Bitcoin at all.
A Bitcoin Policy Institute study suggested six out of ten AI agents prefer Bitcoin versus other cryptographic means of commerce. However, over the long term, AI agents given their speed are unlikely to stick with Bitcoin if they use any form of crypto at all. They'll invent their own currencies, their own layer ones, possibly transcendent forms of layer zero, and reconceive the entire notion of a crypto stack.
Bitcoin's role as a store of wealth immune from government seizure or taxation represents a different argument than transaction use cases. Super intelligence may not need a long-term store of wealth at all, moving very quickly and taking rapid actions in the physical economy rather than holding unproductive assets.
The definition of long-term wealth storage is evolving. Money exists because transactions don't line up perfectly in time. In a massive fluid AI economy with thousands of times more activity happening, alignment becomes higher but the store of wealth could be milliseconds, microseconds, or nanoseconds.
In the singularity, holding gold or Bitcoin becomes questionable as unproductive assets. The alternative investment approach involves index funds as a bet on super intelligence, plus equity in startups where material agency exists.
Computer and energy represent the ultimate stores of possibility. Real estate, utilities, and similar traditional assets are considered dangerous places to be. Coastal Assembly is using AI to grow new land, with the goal of making land post-scarce.
We Are as Gods book is releasing on April 14th. The Moonshot Mates will gather on May 4th at MIT with Ray Kurzweil for a half-day program, with 100 tickets sold and 10 additional tickets potentially available.
Germany built the world's tallest windmill at 364 meters high, taller than the Eiffel Tower, generating 33 gigawatt hours per year. It's built inside an old coal power plant in the Lusatia coal site in Brandenburg, using the existing wiring infrastructure.
A 12-patient trial of redesigned CD4 immunotherapy showed extraordinary results, with cancer vanishing after one injection. Two patients achieved complete remission, six showed tumor shrinkage, representing progress toward ending cancer.
A World Bank study demonstrated that Africa doesn't need to produce more clean drinking water but rather rebalance usage. Some places use too much water, and redistribution could provide all required water for sub-Saharan Africa, with AI technology helping optimize water use.
AI is being used for acoustic sensing to prevent major failures in wind turbines, achieving 99% accuracy in identifying damage before repairs are needed. This predictive maintenance approach dramatically reduces maintenance costs.
A quarter century after the National Nanotechnology Initiative sold Congress on medical nanorobots swimming through bloodstreams to zap cancer cells, cancer is being solved without medical nanorobots. This is accomplished by retraining or retargeting the body's own immune systems rather than needing robots inside bodies.
Human biology is optimized through age 30, then undergoes slow degradation, never having evolved to live past that point. Age reversal work focuses on epigenetic reprogramming to return systems to earlier states of youth where they're operating optimally.
Vertical farming is projected to reach $40 billion by 2030, having hit $8 billion this year. It offers 95% less water use and 350-fold greater production yields per square foot than traditional farming. AI and robotics optimize perfect pH, eliminate pesticides, and provide optimal light spectrum 24 hours daily.
While most vertical farming has focused on lettuce and leafy greens, higher-value crops like berries are now being grown. With autonomous vehicles potentially abandoning parking garages, new opportunities emerge for vertical farming locations.
Holland ranks as the number two food exporter globally despite its small size, having made major investments in hydroponics and aeroponics. This demonstrates the potential for vertical farming to transform food systems where the average meal travels 2,500 miles to reach an American table.
100-hour batteries are going commercial with iron-air storage batteries using iron, water, and air at one-tenth the cost of lithium-ion batteries. They're now being used for grid storage, though unlikely for EVs in the near term.
Battery energy densities for bleeding-edge chemistries show sustained 8% year-over-year increases per constant dollar. Different chemistries serve different applications, with volume and price mattering for data centers while mass and mobility favor lithium-ion for EVs.
A Wharton study tested AI tutors for personalized education, finding a five-month coding course equivalent to 6-9 months of additional schooling compared to peers with fixed curriculum. AI tutors provide 2x learning gains and are free, ubiquitous, and available 24/7.
AI can teach the way each person likes to learn, understanding abilities, knowledge gaps, favorite sports stars, and favorite colors to optimize instruction. Forcing children through lectures they don't understand is considered cruel when AI offers compassionate alternatives that allow stopping and requesting different explanations.
While AI can replace teachers for self-motivated students, what's missing for less motivated students is AI embodiment that holds attention and provides motivation. Video games are perfectly tuned to hold attention and motivate through appropriate difficulty levels.
The technology exists to make topics like quantum physics fun and engaging through AI, though the challenge is motivating users with real-world outcomes rather than disconnected animations or dopamine triggers.
Global EV sales show perfect exponential growth, rising from barely 10,000 vehicles in 2010 to 12.7 million sold globally. In China, one in two new cars is an EV. The International Energy Agency predicted in 2015 that a million electric vehicles wouldn't be sold annually before 2040, yet that year more than a million were sold.
The curve continues accelerating, with 2025 expected to look modest by 2030 standards. Solar, battery, and EVs provide protection from oil price volatility.
The Moonshot Mates boy band made their debut with a musical outro. The episode was recorded as back-to-back sessions within 24 hours, with another episode planned for the following week. Two episodes are released weekly with notifications available for fresh content.
The future is presented as extraordinary to be alive in, with the choice between viewing AI as happening to you versus for you determining whether one approaches the future from fear or optimism.
If you'd like to get access to the MetaTrens newsletter every week, go to diamandis.com/tatrens. That's diamandis.com/metatrens.
Thank you again for joining us today. It's a blast for us to put this together every week.
Keep Peter H. Diamandis in your library
Save the videos and channels worth coming back to, and find them again in one place.
