Latest / Elon Musk Podcast / Anthropic beats OpenAI with enterprise revenue

Transcript

Anthropic beats OpenAI with enterprise revenue · Elon Musk Podcast

• 0:00
  Anthropic has just surpassed Open AI and annualized revenue,
• 0:04
  pulling in $30 billion to Open A is 24 billion.
• 0:08
  Yeah, which is just, it's incredibly revealing when you
• 0:11
  look at the underlying mechanics of how they got there.
• 0:14
  Anthropic achieved this massive revenue figure, relying almost
• 0:17
  entirely on enterprise clients. They completely bypassed the
• 0:20
  consumer market. And you know they did it with
• 0:23
  only a fraction of the user base that Open AI currently supports.
• 0:27
  How did the creator of the most famous artificial intelligence
• 0:30
  product in the world find itself trailing a major competitor,
• 0:35
  fighting an internal executive war and floating the idea of a
• 0:39
  government financial backstop? To figure that out, we really
• 0:42
  have to look closely at the stark contrast and user numbers
• 0:44
  and, well, monetization strategies between the two
• 0:47
  companies. So Entropic is currently
• 0:49
  operating with roughly 134 million active users.
• 0:52
  Open AI, on the other hand, has approximately 900 million active
• 0:56
  users interacting with their systems.
• 0:58
  So when you lay those numbers out side by side, the math on
• 1:02
  the revenue per user gets wild. If you calculate the average
• 1:06
  monthly revenue per active user, Anthropic generates $16.20 per
• 1:11
  person. Wow, yeah.
• 1:13
  Right. And Open AI brings in only $2.20
• 1:15
  per person. Hold on, wait back up.
• 1:18
  How exactly is a company with a fraction of the user base
• 1:21
  extracting that much more money? Uh, person.
• 1:24
  Well, it comes down to a highly targeted enterrise strategy.
• 1:27
  Anthropic is fundamentally ignoring the casual Internet
• 1:30
  user. 80% of their revenue comes directly from corporate clients.
• 1:35
  I see. Yeah, If you look at their
• 1:36
  flagship programming tool, Claude Code, it currently holds
• 1:40
  a commanding 54% market share in the developer segment.
• 1:43
  That single tool loan generates over 2 1/2 billion dollars.
• 1:47
  Right, and Open AI built their entire brand on individual
• 1:49
  consumers signing up for a free chat interface and they are
• 1:54
  still heavily reliant on that free user base.
• 1:56
  Yeah, and the difference in those two user experiences
• 1:59
  completely explains the revenue gap.
• 2:01
  A consumer using a chat application is just incredibly
• 2:04
  fickle. Oh, absolutely.
• 2:05
  They might use it to generate a recipe on a Tuesday, write a
• 2:08
  funny poem for a birthday party on a Thursday, and then, you
• 2:11
  know, not open the application again for three weeks.
• 2:14
  Exactly. They are very unlikely to pay a
• 2:17
  premium monthly subscription for that kind of sporadic utility,
• 2:21
  and even if they do pay, they can cancel their subscription
• 2:24
  and switch to a competitor's application the second they see
• 2:27
  a slightly better feature on social media.
• 2:29
  The switching costs for a consumer are basically 0.
• 2:33
  But a corporate engineering team operates in a completely
• 2:36
  different reality. Yes, exactly.
• 2:38
  When a business integrates a tool like Claude Code into its
• 2:41
  daily engineering workflow, it gets deeply embedded into the
• 2:45
  company's infrastructure. We are talking about integrating
• 2:48
  the software directly into the integrated development
• 2:51
  environment, where programmers write their daily code.
• 2:55
  The software learns the company's specific code base, it
• 2:58
  understands their internal coding guidelines, and it
• 3:00
  becomes a mandatory tool for reviewing code before it gets
• 3:03
  pushed to production. Once that happens, ripping that
• 3:06
  tool out becomes an absolute nightmare.
• 3:08
  It genuinely does. If a chief technology officer
• 3:12
  decides to switch to a different AI provider just to save a few
• 3:15
  dollars a month, they risk breaking the productivity of
• 3:18
  their entire engineering department.
• 3:20
  Because the engineers have to learn a new tool, the system has
• 3:23
  to ingest the entire code base all over again, and thousands of
• 3:27
  hours of workflow optimization are lost.
• 3:29
  Exactly. That makes the corporate
• 3:30
  contract incredibly sticky. Companies will gladly pay large
• 3:35
  recurring, high margin licensing fees because the tool is
• 3:38
  directly generating economic value for them by making their
• 3:42
  highly paid software engineers faster.
• 3:44
  This. Completely up ends the assumed
• 3:46
  business model for the entire industry.
• 3:49
  It establishes that chasing consumer scale does not
• 3:52
  automatically equal revenue scale.
• 3:54
  It diminishes the perceived value of simply hoarding
• 3:57
  hundreds of millions of free users just for the sake of
• 3:59
  market share. And it carves out a much more
• 4:02
  lucrative path entirely on specialized high stickiness
• 4:05
  corporate infrastructure. But earning all that money is
• 4:08
  only 1/2 of the financial equation.
• 4:11
  The other half is the staggering cost of actually training and
• 4:14
  operating the models that power these tools.
• 4:16
  Oh, the financial projections on that side of the business are
• 4:19
  terrifying. Based on current operational run
• 4:22
  rates, Open AI is expected to lose $74 billion in a single
• 4:27
  future reporting period. Wait. 74 billion in one period.
• 4:31
  Yes, and that happens to be the exact same period where
• 4:34
  Anthropic is projected to cross the threshold and achieve actual
• 4:38
  profitability. That kind of disparity doesn't
• 4:41
  happen by accident. Yeah, Anthropic is operating
• 4:43
  with extreme mechanical efficiency.
• 4:45
  They actually spend 4 times less on model training than open AI.
• 4:49
  Wow. And the large part of that comes
• 4:51
  down to how they procure the physical computing required to
• 4:54
  do the work. Anthroic just secured a highly
• 4:57
  flexible 3 1/2 GW compute capacity agreement, right?
• 5:01
  And they are deliberately utilizing multile hardware types
• 5:04
  across Broadcom and Google rather than locking themselves
• 5:07
  into one single supplier. Let's contextualize 3 1/2
• 5:10
  gigawatts because it is hard to conceptualize electricity at
• 5:13
  that volume. A single GW is roughly enough
• 5:16
  electricity to power a medium sized city of about 300,000
• 5:19
  homes. OK, so Anthropics compute
• 5:21
  agreement is drawing enough raw electrical power to run the
• 5:24
  equivalent of the entire metropolitan area of Seattle
• 5:28
  continuously 24 hours a day. That is wild, but the
• 5:32
  flexibility of that power draw is the important part.
• 5:36
  By spreading their infrastructure across Broadcom
• 5:38
  chips and Google's custom hardware, Anthropic is hedging
• 5:42
  their bets. Hedging against what exactly?
• 5:44
  Well, if there's a supply chain shock that makes one type of
• 5:47
  computer chip impossible to buy, or if a geopolitical event
• 5:51
  restricts manufacturing, Anthropic can simply route their
• 5:54
  processing needs through their other hardware partners.
• 5:58
  They aren't tied to a single point of failure.
• 6:00
  Whereas Open AI is pursuing an entirely opposing strategy, they
• 6:04
  are legally locking themselves into a $1.4 trillion
• 6:08
  infrastructure commitment spread over several years.
• 6:10
  Right. And within that massive figure,
• 6:12
  over $120 billion is projected for compute spending alone in
• 6:17
  one single future period. They are committing astronomical
• 6:20
  sums to build out dedicated data centers from the ground up,
• 6:23
  requiring specialized chips from a highly concentrated supply
• 6:26
  chain. Because building a custom data
• 6:28
  center is not like renting server space.
• 6:31
  You have to buy the physical land, secure local government
• 6:34
  permits, build massive cooling systems that consume millions of
• 6:37
  gallons of water. And sign 20 year contracts with
• 6:41
  local utility companies just to ensure you have the electricity
• 6:44
  required to turn the machines on.
• 6:45
  Exactly. Once you start pouring concrete,
• 6:47
  you cannot easily back out. Let's just look at that simply
• 6:50
  one company is buying insurance across different hardware
• 6:53
  platforms to reach profitability, while the other
• 6:56
  is betting everything on a trillion dollar credit card.
• 7:00
  This massive infrastructure burden fundamentally alters open
• 7:04
  a eyes financial flexibility. It restricts their ability to
• 7:07
  pivot or absorb sudden market shocks if consumer demand drops.
• 7:11
  It pushes them into a high stakes, all or nothing race
• 7:13
  where they absolutely must achieve artificial general
• 7:16
  intelligence just to justify the historical costs.
• 7:18
  Right, while freeing Anthropic to operate sustainably, even if
• 7:22
  technological breakthroughs in the broader industry start to
• 7:25
  slow down. So that terrifying cash burn
• 7:27
  rate is creating deep fractures inside Open AI's own executive
• 7:32
  suite. Oh, absolutely.
• 7:34
  There is a severe internal clash happening right now between Open
• 7:39
  AI chief executive Officer Sam Altman and Chief Financial
• 7:42
  Officer Sarah Fryer. Yeah, Altman is aggressively
• 7:44
  pushing for an initial public offering to beat Anthropic to
• 7:47
  the public markets and secure a fresh influx of capital.
• 7:51
  And Fryer is firmly opposing that plan.
• 7:53
  Why is she so opposed? Well, if you want to understand
• 7:57
  the friction in that room, you have to look at Sarah Fryer's
• 8:00
  professional background. She has a reputation for
• 8:03
  ruthless financial discipline. She previously serves as the
• 8:06
  chief financial officer at Square and at Next Door.
• 8:10
  When she was at Next Door, the company was burning through cash
• 8:12
  and missing their profitability targets.
• 8:15
  Fryer oversaw a grisly round of layoffs that cut 25% of the full
• 8:19
  time staff. Wow.
• 8:21
  That was nearly 200 people let go simply to align their
• 8:24
  operating expenses and reach free cash flow amidst millions
• 8:27
  in quarterly losses. Wait, so the person hired
• 8:30
  specifically to prepare them for the public markets is the one
• 8:32
  telling the board they aren't ready?
• 8:34
  Exactly. The chief financial officers
• 8:37
  primary job when preparing for an initial public offering is to
• 8:41
  ensure the company can withstand the scrutiny of institutional
• 8:45
  investors and federal regulators.
• 8:47
  Right. When a company files an S1
• 8:49
  document to go public, they have to open their books completely.
• 8:53
  Every single liability, every ongoing contract and every
• 8:57
  potential risk has to be disclosed to the public.
• 9:00
  Fryer is issuing clear warnings internally that the company
• 9:03
  simply cannot meet the rigorous reporting standards required of
• 9:07
  public companies given their unsustainable server leasing
• 9:10
  commitments. Because when you lease server
• 9:12
  space for billions of dollars, those leases sit on your balance
• 9:16
  sheet as massive liabilities. Public market investors hate
• 9:19
  seeing a balance sheet weighed down by fixed costs that extend
• 9:22
  years into the future, especially when the revenue tied
• 9:25
  to those costs is highly unpredictable.
• 9:27
  The tension over this reality has become so severe that Fryer
• 9:30
  has actually been excluded from key investor meetings regarding
• 9:33
  server spending. The CEO is holding high level
• 9:35
  capital expenditure discussions without the Chief Financial
• 9:38
  Officer in the room. That kind of governance in its
• 9:41
  chaos leaks out into the broader financial ecosystem.
• 9:45
  You can see the impact directly and how the company shares are
• 9:47
  being valued. Currently, open AI shares on the
• 9:50
  secondary market are trading at a 10% discount.
• 9:54
  Meanwhile, Anthropic shares are commanding a 50% premium.
• 9:58
  The secondary market is essentially where employees and
• 10:01
  early investors sell their private shares to outside
• 10:03
  institutions before a company goes public.
• 10:06
  It is the purest indicator of what Wall Street actually thinks
• 10:09
  a private company is worth. Exactly.
• 10:11
  If institutional buyers are demanding a 10% discount on open
• 10:15
  AI shares, it means they believe the company's official valuation
• 10:19
  is inflated. This internal chaos degrades how
• 10:21
  institutional investors view the company.
• 10:24
  It limits Open AI's ability to project unified leadership to
• 10:27
  the public markets, and it exposes severe vulnerabilities
• 10:30
  that competitors are actively exploiting in private funding
• 10:33
  rounds. Capital always flows towards
• 10:35
  stability and right now the secondary market pricing
• 10:38
  reflects a strong preference for Anthropics disciplined,
• 10:41
  financially rigorous approach. And the financial strain
• 10:45
  highlighted by Fryer is already forcing Open AI to make drastic
• 10:48
  operational sacrifices. Open AI recently shut down its
• 10:51
  highly anticipated video generation model, SORA, directly
• 10:55
  due to compute allocation constraints.
• 10:57
  They literally ran out of the necessary processing power to
• 11:00
  keep the project alive. The financial weight of that
• 11:03
  specific project was immense. Generating video is an entirely
• 11:07
  different beast than generating text.
• 11:09
  Sora was consuming approximately $15 million worth of compute
• 11:12
  resources every single day. Every single day.
• 11:16
  To understand why it cost so much, you have to look at the
• 11:19
  physics of rendering a video file.
• 11:21
  Right When a user asks a text model to write an essay, the
• 11:23
  computer is calculating the statistical probability of the
• 11:26
  next word, stringing together text characters.
• 11:29
  It requires compute power, but it is, you know, relatively
• 11:32
  lightweight. Video generation is
• 11:34
  exponentially more demanding. A single second of high
• 11:37
  definition video contains 30 to 60 individual high resolution
• 11:41
  images or frames. So for a 10 second video, the
• 11:44
  system isn't just generating one response, it is rendering 600
• 11:48
  highly detailed images. And then it has to calculate the
• 11:52
  physics, lighting and consistency between every single
• 11:56
  one of those frames so the motion looks natural.
• 11:58
  Doing that for millions of users simultaneously burns through an
• 12:02
  unbelievable amount of electricity and processor
• 12:04
  capacity. The fundamental problem for Open
• 12:07
  AI was at the exact same batch batch of specialized hardware
• 12:11
  needed to run the sore video model was also needed to train
• 12:14
  their flagship large language model GPT 6.
• 12:17
  They could not do both right. Altman had to make a choice
• 12:20
  between the two, and the video model was axed so they could
• 12:23
  focus their finite resources entirely on the core text model.
• 12:26
  I see that as a massive strategic mistake.
• 12:29
  Shutting down sore essentially concedes the entire video
• 12:32
  generation market to competitors, right as the
• 12:35
  underlying technology is maturing.
• 12:37
  Video is the dominant medium of the modern Internet.
• 12:39
  Well, I completely disagree with the assessment.
• 12:41
  I think it was a totally necessary survival tactic.
• 12:44
  But TikTok and YouTube command the vast majority of consumer
• 12:47
  attention. Walking away from video gives up
• 12:50
  an enormous surface area for future consumer engagement and
• 12:54
  leaves a multibillion dollar advertising and entertainment
• 12:57
  market sitting on the table operating.
• 12:59
  A short form video feed factory is a low margin distraction.
• 13:03
  Generating video requires massive amounts of processing
• 13:06
  power for outbound that is incredibly difficult to monetize
• 13:09
  effectively. You don't think consumers will
• 13:11
  pay for it? How much will a consumer
• 13:13
  actually pay to generate a weird 10 second video of a cat flying
• 13:17
  a spaceship? Very little.
• 13:20
  It distracts from the pursuit of a super app strategy centered
• 13:23
  around high value text based reasoning and complex enterprise
• 13:27
  agents. If you don't secure the core
• 13:29
  reasoning product first, the flashy video features won't save
• 13:32
  the business. Regardless of the underlying
• 13:34
  strategy, this shatters the perception of infinite compute.
• 13:38
  It strictly limits Open AI's ability to monopolize every
• 13:41
  single sector of the industry simultaneously, and it creates a
• 13:44
  wide open vacuum for specialized startups to claim the video
• 13:47
  generation space without facing an unstoppable, well funded
• 13:50
  incumbent. And while Open AI is forced to
• 13:53
  ration its computing power and pick its battles, Alphabet is
• 13:56
  showing exactly what happens when a company owns the entire
• 14:00
  supply chain. Alphabet's recent earnings
• 14:03
  report is staggering. They reported over $109 billion
• 14:08
  in consolidated revenue. A major highlight in that report
• 14:12
  was a 63% surge in Google Cloud revenue, bringing that single
• 14:16
  division to $20 billion. The specific drivers behind that
• 14:21
  surge are incredible to look at. Cloud Generative AI model
• 14:24
  revenue grew by nearly 800%. 800%.
• 14:27
  Yeah, and paid monthly active users for their Gemini
• 14:29
  enterprise solutions jumped 40%. They are seeing massive adoption
• 14:34
  from corporate clients who are more than willing to pay premium
• 14:37
  rates for integrated services. Alphabet occupies a truly unique
• 14:40
  position in this industry. They are not just selling
• 14:43
  software subscriptions, they are actually selling their 8th
• 14:45
  generation TPU hardware directly to select customers to put into
• 14:50
  their own data centers. Which creates an entirely new
• 14:53
  revenue stream that directly challenges traditional chip
• 14:55
  makers like NVIDIA. For anyone listening, a TPU is a
• 14:59
  Tensor Processing unit. It is a custom designed piece of
• 15:02
  hardware built specifically by Google to accelerate artificial
• 15:05
  intelligence workloads. Most companies have to buy
• 15:08
  graphics processing units or GPU's from outside vendors.
• 15:13
  Because Alphabet designs and builds their own Tpus, they
• 15:16
  bypass the incredible markup that third party hardware
• 15:19
  manufacturers charge. And now, by selling those Tpus
• 15:23
  directly to other companies, they are monetizing the physical
• 15:26
  infrastructure layer itself. Meanwhile, their core search
• 15:29
  revenue continues to grow robustly, hitting over $60
• 15:32
  billion. Right alongside half a million
• 15:35
  weekly autonomous rides from Waymo showing they're
• 15:38
  successfully deploying these models into the physical world.
• 15:41
  To help you understand this dynamic Alphabet is like a
• 15:43
  casino that owns the slot machines, the hotel and the
• 15:46
  power plant supplying the electricity.
• 15:48
  That is a great way to put it, and Open AI is just a High
• 15:50
  Roller renting a very expensive suite inside that hotel.
• 15:54
  Alphabet controls the foundational infrastructure, the
• 15:56
  hardware design, the software layer, and the distribution
• 15:59
  channels across billions of Android phones in Chrome
• 16:02
  browsers. This vertical integration
• 16:05
  fundamentally alters the risk profile for investors.
• 16:08
  It positions Alphabet to capture massive profits from the
• 16:11
  infrastructure layer itself, drastically reducing their
• 16:14
  exposure if any single consumer facing chat product fails to
• 16:19
  gain traction. Yeah, they went as long as
• 16:21
  anyone, anywhere requires computing power.
• 16:24
  Alphabet self-contained, highly profitable ecosystem stands in
• 16:28
  stark contrast to the rest of the industry, which is currently
• 16:31
  propped up by a highly complex, interconnected web of debt.
• 16:34
  This brings us to the concept of the circular investment loop,
• 16:37
  where the money never actually leaves the room.
• 16:39
  The exact flow of this loop is incredibly intricate.
• 16:42
  Let's trace the money here. Open AI commits hundreds of
• 16:45
  billions of dollars to Oracle for cloud computing power.
• 16:48
  OK, then Oracle borrows money from Wall Street to build new
• 16:51
  physical data centers and buy NVIDIA hardware to fulfill that
• 16:55
  specific contract. Right.
• 16:57
  NVIDIA takes a portion of that revenue and buys equity in
• 17:00
  smaller cloud providers like a company called Cor Weave.
• 17:03
  NVIDIA also pre purchases services from Cor Weave,
• 17:06
  injecting them with cash so Cor Weave can confidently go to a
• 17:09
  bank, get a loan and buy even more NVIDIA hardware.
• 17:13
  And Cor, Weave's biggest customer happens to be
• 17:15
  Microsoft, and Microsoft is the primary funder and largest
• 17:19
  shareholder of Open AI. Let's.
• 17:21
  Simplify this for you. Imagine me loaning you $100 so
• 17:24
  you can buy lemonade from my kids stand.
• 17:26
  OK, I follow. Then my kid takes that same $100
• 17:30
  and pays me rent for using the front yard.
• 17:32
  On paper, our family just generated $200 in economic
• 17:35
  activity. We can show investors A booming
• 17:38
  lemonade business and a lucrative real estate business.
• 17:41
  But in reality, no new money entered the room.
• 17:43
  Exactly, it's just a group of companies passing the exact same
• 17:47
  dollar bill around in a circle and every single time it changes
• 17:50
  hands they recorded on their balance sheets as brand new
• 17:54
  organic revenue. The major players are starting
• 17:57
  to realize the profound danger of the structure.
• 17:59
  NVIDIA recently backed down from a highly publicized $100 billion
• 18:04
  investment in Open AI. They quietly settled for just 30
• 18:08
  billion and the CEO confirmed it will likely be their last
• 18:12
  private investment in these specific research labs.
• 18:16
  They are actively reducing their exposure to this exact loop.
• 18:20
  This interconnected structure severely threatens the overall
• 18:23
  stability of the tech sector. It masks true organic market
• 18:27
  validation because the revenue is generated almost entirely by
• 18:31
  companies funding each other. It exposes the entire industry
• 18:34
  to severe contagion if one link in that chain breaks.
• 18:37
  If a company cannot service its massive debt load or fulfill a
• 18:40
  promised compute contract, the entire structure could collapse
• 18:44
  inward on itself. The fear of that exact collapse
• 18:46
  is driving some executives to seek a permanent safety net,
• 18:49
  leading them directly to the federal government.
• 18:51
  Open AI executives, including the CEO and the CFO, have
• 18:55
  publicly floated the idea of securing federal loan guarantees
• 18:58
  or a government financial backstop to help fund their
• 19:00
  trillion dollar infrastructure push.
• 19:02
  The backlash to this idea was immediate and fierce.
• 19:06
  Consumer advocacy groups and financial critics immediately
• 19:09
  labeled this a bailout for a speculative bubble.
• 19:12
  They argued forcefully that tech companies want to socialize the
• 19:16
  massive risks of their infrastructure build out while
• 19:19
  keeping the eventual profits entirely private.
• 19:22
  Because a federal loan guarantee means that if the company
• 19:24
  defaults on its trillion dollar debt, the American taxpayer is
• 19:28
  legally obligated to step in and pay the bill.
• 19:31
  There is a much broader economic context here.
• 19:34
  Financial analysts are calling this the rot economy, where the
• 19:37
  entire focus is strictly on growth at all costs without any
• 19:41
  underlying sustainable utility. They draw direct comparisons to
• 19:45
  the.com bubbles dark fiber phenomenon, the.
• 19:48
  Dark fiber comparison is incredibly relevant.
• 19:51
  During the height of the.com boom in the late 90s,
• 19:53
  telecommunications companies convinced themselves that
• 19:56
  Internet traffic was going to grow exponentially forever.
• 19:59
  So they borrowed billions of dollars to lay thousands of
• 20:01
  miles of highly expensive fiber optic cables across the country
• 20:05
  and under the oceans. But the consumer demand just
• 20:08
  wasn't there yet. When the bubble burst, those
• 20:11
  companies went bankrupt and miles of cable sat completely
• 20:14
  unused, entirely dark for years. We are seeing a similar
• 20:18
  disconnect today. Currently, 95% of enterprise
• 20:22
  implementations fail to show a return on investment, yet
• 20:26
  companies are continuously issuing hundreds of billions in
• 20:29
  corporate debt to build massive new data centers.
• 20:32
  You have to consider that government backing might be
• 20:35
  entirely necessary for national security, though staying ahead
• 20:38
  of global competitors and technological infrastructure is
• 20:41
  critical. I am not sure I agree with that.
• 20:44
  If foreign adversaries build these massive data centers 1st
• 20:47
  and achieve technological superiority, the economic and
• 20:50
  military consequences could be devastating.
• 20:52
  Private capital markets might simply not be capable of
• 20:55
  sustaining the required investment levels for these
• 20:58
  fundamental nation state level technological leaps.
• 21:01
  I push back on that hard. The absurdity of taxpayers
• 21:04
  subsidizing a private company with a valuation approaching a
• 21:07
  trillion dollars is clear. They overextended themselves on
• 21:10
  speculative hardware commitments and now they want a safety net.
• 21:13
  I mean, the geopolitical stakes are real.
• 21:16
  Yes, but if the business model is actually sound and the
• 21:19
  technology delivers the value it promises, private capital will
• 21:22
  fund it. If it isn't sound, the public
• 21:24
  shouldn't carry the risk for a private entities massive gamble
• 21:28
  under the guise of national security.
• 21:30
  This request for loan guarantees fundamentally shifts the
• 21:33
  political conversation around the technology.
• 21:36
  It strips away the narrative of unstoppable organic innovation,
• 21:39
  and it invites serious regulatory scrutiny regarding
• 21:42
  whether the current infrastructure build out is
• 21:45
  actually driven by real world demand or just speculative
• 21:48
  momentum fueled by cheap debt. The race to build artificial
• 21:51
  intelligence has fundamentally shifted from who has the
• 21:54
  smartest software to who has the most sustainable business model.
• 21:58
  The staggering costs of infrastructure and the fragility
• 22:01
  of circular investments are forcing companies to either find
• 22:04
  real enterprise revenue like Ampropic, or look for outside
• 22:08
  bailouts. Music finally stops and this
• 22:10
  trillion dollar infrastructure bill comes due.
• 22:12
  Who will actually be left holding the check?
• 22:15
  If you're not subscribed yet, take a second and hit follow on
• 22:18
  whatever app you're using. It helps us keep making this.
• 22:20
  We appreciate you being here. Also, check out our YouTube
• 22:23
  channel for more business and tech updates.
• 22:25
  There's a link in the description.