Latest / Elon Musk Podcast / Anthropic surpasses OpenAI despite military ban

Transcript

Anthropic surpasses OpenAI despite military ban · Elon Musk Podcast

• 0:00
  Anthropic has officially crossed a $30 billion annualized revenue
• 0:05
  run rate, officially surpassing Open AI in revenue generation.
• 0:09
  Yeah, the velocity of this financial growth is just
• 0:12
  genuinely difficult to comprehend, right, especially
• 0:15
  considering this company was generating just a fraction of
• 0:17
  that amount in the very recent past.
• 0:19
  And to put this in perspective for you, secondary markets are
• 0:23
  currently placing an implied valuation of $1 trillion on the
• 0:27
  company A. Trillion dollars.
• 0:28
  Exactly. And they are achieving all of
• 0:31
  this while simultaneously fighting a ban from the United
• 0:34
  States military. Which is just crazy.
• 0:36
  I mean, how did a research lab that didn't even exist a few
• 0:39
  years ago managed to overtake the most recognizable name in
• 0:43
  artificial intelligence, all while becoming the center of
• 0:46
  this huge geopolitical standoff? Well, we are basically looking
• 0:49
  at a complete realignment of the artificial intelligence industry
• 0:51
  here. The focus has heavily shifted
• 0:53
  toward their enterprise strategy, the unprecedented
• 0:57
  physical infrastructure build out required to maintain it, and
• 1:00
  the release of a highly restricted eponymous
• 1:02
  cybersecurity model, Right, right.
• 1:04
  And of course, that resulting friction with the Pentagon over
• 1:07
  who actually controls the deployment of this technology.
• 1:10
  So we have to look at the vertical climb of Anthropics
• 1:12
  revenue to really understand the mechanics at play here.
• 1:15
  The annualized recurring revenue went from $1 billion to 9
• 1:20
  billion to 14 billion and recently hit that $30 billion
• 1:24
  mark. It's wild.
• 1:26
  It is, and you compare that to open AI is reported $25 billion
• 1:31
  and the contrast and trajectory becomes incredibly sharp.
• 1:35
  Wait, back up. How exactly are these two
• 1:37
  companies counting their money? Good question.
• 1:39
  Because $30 billion is a massive number, but accounting practices
• 1:43
  in the cloud computing sector can vary wildly, right?
• 1:47
  We really need to be precise about what that number actually
• 1:49
  represents. Yeah, that.
• 1:50
  Is a crucial distinction. The accounting differences
• 1:53
  explain a significant part of how these revenues are
• 1:55
  calculated. Anthropic reports Gross
• 1:58
  Annualized Recurring Revenue, or ARR.
• 2:01
  That means they count the full cloud sale 1st and then they pay
• 2:05
  the cloud provider their cut. But Open AI reports net ARR,
• 2:09
  meaning they only count the revenue after the partner like
• 2:12
  Microsoft takes their cut. Oh, I see.
• 2:14
  So if your company spends $10 on an Anthropic product through
• 2:18
  Amazon Web Services, Anthropic counts the full $10 as revenue,
• 2:23
  then pays Amazon their infrastructure fee, right?
• 2:26
  If you spend $10 on an Open AI product through Azure, Open AI
• 2:30
  might only record the $7.00 they actually keep.
• 2:33
  The counting mechanisms differ entirely.
• 2:35
  We can extend that point by looking at their fundamental
• 2:38
  business models, actually, which sit at opposite ends of the
• 2:41
  spectrum. Yeah, Open AI relies heavily on
• 2:43
  consumer subscriptions. They have hundreds of millions
• 2:47
  of weekly active users, but those individual users are
• 2:49
  paying a small monthly fee. Right.
• 2:51
  Like 20 bucks or whatever. Exactly.
• 2:53
  And the conversion rate from a free user to a paid user is
• 2:56
  remarkably low, sitting around 5%.
• 2:58
  Wow. Just 5%.
• 3:00
  Meanwhile, Anthropic derives 80% of its revenue directly from
• 3:04
  enterprise clients and API calls.
• 3:06
  Huge difference. It really is.
• 3:08
  The metric that highlights the difference in these strategies
• 3:10
  is revenue per user. Anthropic pulls in roughly 211
• 3:15
  dollars per monthly user. Oh wow.
• 3:17
  Yeah, compared to Open AI's $25 per weekly user.
• 3:21
  You could really see the success of this enterprise focus in the
• 3:24
  specific deployment of products like Claude Code and Claude
• 3:29
  Cowork. Oh, definitely.
• 3:30
  Claude code reached a 2 1/2 billion dollar run rate in a
• 3:33
  matter of months. Businesses are not just using
• 3:36
  this to draft emails, you know, right?
• 3:39
  They're using it to automate incredibly complex coding tasks
• 3:43
  within their internal environments.
• 3:45
  We are looking at a scenario where eight of the Fortune 10
• 3:48
  companies use Claude. Eight of the top ten?
• 3:51
  That's wild. Yeah, and 4% of all public
• 3:54
  GitHub commits globally are authored by Claude code.
• 3:57
  Wait, 4% of all of? Them, all of them, globally.
• 4:00
  And to understand the gravity of that for you, GitHub is the
• 4:03
  repository where the vast majority of the world's open
• 4:06
  source software is stored and updated.
• 4:09
  If 4% of all changes to that global code base are being
• 4:13
  written by an autonomous system, you are looking at a massive
• 4:17
  footprint in the foundational architecture of modern software.
• 4:21
  Well, this completely limits the commercial ceiling for consumer
• 4:24
  chat bots. How so?
• 4:26
  Consumer applications suffer from really high churn rates.
• 4:30
  If an individual gets bored or finds a cheaper alternative,
• 4:34
  they cancel their $20 subscription instantly.
• 4:37
  Oh for sure, very easy to just cancel, right?
• 4:40
  But what Anthropics model opens up is a highly lucrative sticky
• 4:44
  market for autonomous enterprise workflow integration.
• 4:48
  Sticky is the keyword there. Exactly.
• 4:50
  When a multinational corporation integrates an AI model into
• 4:54
  their proprietary code base and daily operations, ripping that
• 4:57
  infrastructure out and replacing it takes years.
• 5:00
  Nobody wants to. Do that, the revenue basically
• 5:01
  becomes locked in. It's the difference between
• 5:04
  selling a single ticket to 1,000,000 tourists versus
• 5:07
  securing an exclusive contract to supply the engine parts for a
• 5:10
  massive global airline. That's a great way to put it.
• 5:12
  The airline relies on you to stay in the sky.
• 5:15
  If your product works, the switching costs are simply too
• 5:19
  high for them to ever leave. The secondary market for these
• 5:22
  shares reflects exactly that reality, too.
• 5:24
  Oh yeah. Let's talk about that.
• 5:26
  If we look at the official Series G funding round,
• 5:30
  Anthropic raised $30 billion at a $380 billion post money
• 5:35
  valuation. Which is already a huge number.
• 5:38
  Right, this round was backed by massive sovereign wealth and
• 5:41
  institutional entities like GIC, CO2, MGX, and DE Shaw.
• 5:47
  But you have to contract that official valuation with what is
• 5:50
  actually happening on the secondary market.
• 5:53
  On trading platforms like Forge Global and Jupiter, pre IPO
• 5:57
  instruments imply a $1 trillion valuation.
• 6:00
  A. Trillion.
• 6:01
  Some specific bids are even reaching $1.5 trillion.
• 6:05
  That is just hard to wrap your head around.
• 6:07
  Even on a more conservative trading platform like Hive, the
• 6:10
  valuation sits at $851 billion. The investor behavior
• 6:15
  surrounding these shares is completely frantic right now.
• 6:17
  There is $2 billion in buy side demand chasing shares with
• 6:22
  almost no sellers willing to part with their equity.
• 6:24
  Nobody wants to sell. Exactly, luckily you have
• 6:27
  investors literally offering a 14 acre estate in exchange for
• 6:31
  share. In an entire estate, that's
• 6:33
  wild. Institutional brokers are also
• 6:35
  reflecting this disparity in demand.
• 6:38
  Goldman Sachs is charging a 15 to 20% carry fee for anthropic
• 6:42
  allocations, while simultaneously waving fees
• 6:45
  entirely just to push open AI shares.
• 6:48
  OK, so let's just reset the pace for a second here.
• 6:51
  Yeah, when people who invest for a living are willing to trade
• 6:54
  their houses for a piece of paper, the market market has
• 6:56
  stopped looking at current profits and is betting entirely
• 7:00
  on total global dominance. I mean, I actually completely
• 7:03
  disagree with applying that logic without a serious degree
• 7:06
  of caution. Really.
• 7:07
  Why? A trillion dollar valuation for
• 7:09
  a company that makes $30 billion in revenue is mathematically
• 7:13
  perilous. Well, the multiples are high,
• 7:15
  sure. The multiples are stretched far
• 7:17
  beyond historical precedent for any software or infrastructure
• 7:21
  company in existence. You're basically assuming
• 7:24
  perfect execution for the next decade.
• 7:27
  But the buyers are not pricing in $30 billion.
• 7:30
  True, they're pricing in the company potentially hitting $100
• 7:34
  billion revenue very soon. I guess so.
• 7:36
  They are looking at the trajectory the enterprise lock
• 7:39
  in we just discussed and the fact that the actual cost of
• 7:42
  computation is scaling aggressively alongside the
• 7:45
  revenue. They view the current revenue as
• 7:48
  merely the opening act. Well, this changes the entire
• 7:50
  dynamic of capital allocation regardless.
• 7:53
  Yeah, it opens up a new reality for private markets, where
• 7:56
  secondary trading sets the true price of technology companies
• 8:00
  long before they ever go public. Exactly.
• 8:03
  This fundamentally limits the influence of traditional initial
• 8:06
  public offerings. All the price discovery and all
• 8:09
  the massive valuation leaps are happening behind closed doors.
• 8:12
  Right. Accessible only to accredited
• 8:14
  investors and sovereign wealth funds.
• 8:16
  Yeah, completely locking out the retail investor from
• 8:19
  participating in the largest wealth creation events.
• 8:22
  To maintain that astronomical level of growth, the physical
• 8:26
  requirements are just staggering.
• 8:28
  We have to look at the massive infrastructure investments from
• 8:31
  hyperscalers. At the power stuff.
• 8:33
  Yes, Google committed up to $40 billion, providing 5 gigawatts
• 8:38
  of TPU capacity, and Amazon committed $25 billion, providing
• 8:45
  5 gigawatts of Tranium capacity. These deals are highly
• 8:48
  reciprocal in nature though, yeah.
• 8:50
  They go both ways. Anthropic secured the funding,
• 8:53
  but they simultaneously committed to spending more than
• 8:55
  $100 billion over the next decade on Amazon Web Services
• 8:59
  technologies. This includes a heavy reliance
• 9:01
  on their custom silicon, specifically Tranium 3 and
• 9:04
  Graviton processors. The sources point to their
• 9:07
  collaboration on Project Rainier.
• 9:09
  Project Rainier. Yeah, which is a massive AI
• 9:11
  compute cluster featuring half a million chips working in tandem.
• 9:15
  Hold on, wait. Back up.
• 9:16
  We need to pause and define what 5 gigawatts actually means in
• 9:19
  the real world. Good idea.
• 9:20
  That metric gets thrown around constantly in these
• 9:23
  infrastructure reports, but it is incredibly difficult to
• 9:26
  conceptualize without a direct comparison.
• 9:29
  Yeah. So a single GW can power
• 9:31
  hundreds of thousands of homes. OK.
• 9:34
  When we talk about securing 5 gigawatts from Google and
• 9:37
  another 5 gigawatts from Amazon, the energy required to train
• 9:41
  these models is no longer comparable to running a
• 9:44
  traditional server farm. It's way beyond that.
• 9:47
  It is the equivalent of acquiring the power grid of a
• 9:50
  mid sized country. You were dealing with massive
• 9:52
  land acquisition, dedicated electrical stations, and cooling
• 9:56
  infrastructure on an industrial scale.
• 9:58
  Wow. You're essentially building a
• 9:59
  small city purely to house processors.
• 10:02
  And this fundamentally limits the entire field of competition.
• 10:05
  Absolutely. It limits the AI race to only
• 10:08
  those entities capable of securing multi GW power
• 10:12
  agreements, effectively locking out any new startups from
• 10:15
  competing at the frontier level. Yeah, it's a huge barrier to
• 10:18
  entry. If you have a brilliant
• 10:19
  algorithm, it does not matter if you cannot physically plug it
• 10:23
  into a power source large enough to train it.
• 10:25
  The era of a few engineers building a frontier model in a
• 10:28
  garage is completely over. And to manage the integration of
• 10:32
  these massive models into corporate environments, Entropic
• 10:35
  introduced the Model Context Protocol, or MCP.
• 10:39
  MCP, right? They open sourced this protocol
• 10:41
  and it was later donated to the Linux Foundation.
• 10:44
  The explicitly stated goal was to create a standard way for AI
• 10:48
  agents to connect to external data sources.
• 10:51
  Well, like a. Bridge.
• 10:52
  Exactly. If you want an AI to read your
• 10:54
  customer relationship management system or pull data from your
• 10:57
  secure file systems, MCP provides the standardized
• 11:00
  pathway to do that. We have a highly detailed survey
• 11:03
  data from Zooplo that illustrates exactly how fast
• 11:06
  this protocol is spreading the. Adoption is crazy.
• 11:09
  It is the model context protocol has reached 97,000,000 installs.
• 11:13
  97 million. 70% of users configure between 2:00 and 7:00
• 11:18
  MCP servers for their AI environments, and what is
• 11:22
  fascinating mechanically is that 59% are using streamable HTTP
• 11:27
  for transport rather than the default STDIO.
• 11:31
  OK, wait, what's the difference there?
• 11:32
  Well. STDIO is typically used for
• 11:34
  local machine to machine communication, while streamable
• 11:37
  HTTP allows for a continuous data streaming over networks.
• 11:41
  The heavy use of streamable Http://indicates highly complex
• 11:45
  remote server deployments rather than simple local testing.
• 11:49
  But the survey also reveals significant hurdles accompanying
• 11:52
  this rapid adoption. Oh yeah, the security issues.
• 11:54
  50% of builders site security and access control as their
• 11:57
  absolute biggest challenge. That makes sense.
• 12:00
  Even more concerning from a structural standpoint, 24% of
• 12:03
  MCP servers have no authentication at all.
• 12:06
  None. 0. 0 Authentication To manage this complexity and
• 12:10
  attempt to secure these endpoints, 30% of developers are
• 12:13
  hosting these servers on API gateways.
• 12:16
  An API gateway acts as a traffic cop, basically verifying
• 12:19
  requests before they reach the sensitive data.
• 12:22
  But the sheer volume of unauthenticated servers remains
• 12:25
  A glaring vulnerability. Now Anthropic bills this as an
• 12:28
  open standard to help everyone in the industry communicate
• 12:31
  smoothly. They present it as a rising tide
• 12:35
  lifting all boats. Well, that is the public
• 12:37
  messaging, but I view it as a highly strategic defensive
• 12:41
  maneuver. OK, how so?
• 12:42
  By establishing the protocol that everyone uses, Anthropic
• 12:45
  ensures its ecosystem becomes the permanent infrastructure
• 12:49
  layer of the Internet. Think of it as creating the USBC
• 12:52
  for AI. Oh, I see.
• 12:54
  Even if a competitor builds a slightly better model next year,
• 12:57
  all the corporate data pipes, all the API gateways and all the
• 13:00
  enterprise permissions are already perfectly formatted for
• 13:03
  Anthropic's protocol. The switching cost becomes
• 13:07
  prohibitive. Right.
• 13:08
  This opens up a universal standard for AI tool
• 13:10
  integration, allowing enterprise developers to connect incredibly
• 13:14
  secure databases to large language models seamlessly.
• 13:17
  However, it introduces severe access control risks.
• 13:21
  If the AI agent has the keys to your entire corporate database
• 13:24
  to perform its tasks, and the authentication layer is weak or
• 13:28
  non existent, the entire corporate network becomes
• 13:31
  vulnerable to external exploitation.
• 13:33
  And that exact vulnerability leads directly into the
• 13:35
  development of Mythos Mythos. Mythos is Anthropic's highly
• 13:39
  restricted cybersecurity model. Its capabilities extend far
• 13:43
  beyond standard conversational AI.
• 13:46
  The sources detail how it can complete a 32 step cyberattack
• 13:49
  simulation entirely without human intervention.
• 13:52
  Wow. 32 steps autonomously. Yeah, that is an act that
• 13:56
  normally takes professional, highly trained human hackers
• 13:59
  days to execute as they manually test endpoints, look for
• 14:02
  misconfigurations, and attempt to escalate privileges.
• 14:06
  Mythos automates that entire lateral movement process.
• 14:09
  Mozilla utilized Mythos in a secure environment and found 271
• 14:13
  zero day vulnerabilities in the Firefox browser 2. 171, yes.
• 14:18
  And a zero day vulnerability is a soccer flaw unknown to the
• 14:22
  vendor, meaning there is zero time to fix it before it can be
• 14:24
  exploited. Finding 271 of them autonomously
• 14:28
  is unprecedented. The power of this specific model
• 14:31
  LED directly to the Project Glasswing initiative.
• 14:34
  This program heavily restricted Mythos access to a very select
• 14:39
  group of organizations, primarily focusing on critical
• 14:41
  infrastructure partners like Apple and JP Morgan.
• 14:45
  The objective was to allow these entities to continuously scan
• 14:48
  their own code bases and patch their vulnerabilities long
• 14:52
  before state sponsored adversaries could exploit them.
• 14:55
  Then the breach occurred. Yes.
• 14:58
  The breach, A Bloomberg report demonstrated that unauthorized
• 15:01
  users successfully gained access to the Mythos preview, and the
• 15:05
  method of access was surprisingly rudimentary.
• 15:07
  It wasn't some complex hack. No, not at all.
• 15:10
  A forum user working for a third party contractor combined their
• 15:14
  authorized basic credentials with data obtained from an
• 15:17
  unrelated data breach at the startup.
• 15:19
  Mercker. Oh wow.
• 15:20
  By cross referencing this information, they managed to
• 15:22
  locate and access the restricted model.
• 15:25
  Wait, hold on back up. How was a model explicitly
• 15:29
  designed to secure the most critical financial and
• 15:31
  technological systems in the world compromised through a
• 15:35
  basic supply chain vendor flaw? Exactly.
• 15:38
  Let's just state this plainly. The most advanced AI hacker in
• 15:42
  the world was accessed because a human contractor left the
• 15:46
  digital door unlocked. Yeah, that's exactly what.
• 15:49
  Happened. It wasn't a sophisticated
• 15:50
  algorithmic jailbreak. Yeah, it was basic human error
• 15:54
  and poor access control within the extended vendor network.
• 15:57
  This completely changes the paradigm of cybersecurity
• 16:00
  defense. It opens up the possibility of
• 16:02
  automated 0 day detection where vulnerabilities are found and
• 16:05
  patched instantly shifting the advantage back to the defenders
• 16:09
  right. However, it's severely limits
• 16:10
  the confidence that private companies can safely hoard
• 16:13
  frontier models. If a third party contractor can
• 16:15
  leak access through simple credential reuse, hostile nation
• 16:19
  states certainly possess the capability to do the same.
• 16:21
  Which brings us to the government side of things.
• 16:24
  Before we get into this, we must state clearly for Youth listener
• 16:27
  that we are impartially reporting the facts of the
• 16:29
  dispute between the United States government and the
• 16:32
  company strictly as outlined in our sources, and we are not
• 16:35
  endorsing either political viewpoint.
• 16:37
  Yes, absolutely. This brings us directly to the
• 16:39
  contract dispute with the Pentagon.
• 16:42
  Anthropic secured a $200 million contract to deploy Claude on
• 16:47
  classified government networks. However, they maintained
• 16:51
  incredibly strict acceptable use policies.
• 16:54
  Which caused some friction. Oh, big time.
• 16:58
  These policies explicitly prohibited the use of their
• 17:01
  technology for fully autonomous weapon systems and for mass
• 17:04
  domestic surveillance operations.
• 17:06
  The Department of Defense rejected these specific
• 17:08
  limitations and demanded completely unrestricted access
• 17:12
  for all lawful purposes. And after those contract
• 17:14
  negotiations failed to reach an agreement, the fallout was
• 17:18
  immediate and severe. It really was.
• 17:20
  Defense Secretary Pete Hegseth officially designated Anthropic
• 17:24
  a supply chain risk. The President subsequently
• 17:27
  directed federal agencies to begin a comprehensive phase out
• 17:30
  of anthropic technology across the entirety of the federal
• 17:34
  government. The legal mechanisms utilized to
• 17:36
  execute this phase out are highly specific.
• 17:39
  Actually, the government invoked AFSEA say order, which stands
• 17:44
  for the Federal Acquisition Supply Chain Security Act
• 17:47
  alongside 10 USC Section 3252, right?
• 17:51
  A fast CSA order functions essentially as an emergency kill
• 17:55
  switch for federal procurement, right?
• 17:57
  These mechanisms force government contractors across
• 17:59
  the country to immediately halt new deployments, inventory their
• 18:03
  existing use of clawed, and actively seek alternative
• 18:06
  vendors. Just total chaos for
• 18:07
  contractors. Exactly.
• 18:09
  And right in the middle of this logistical chaos Open AICEO Sam
• 18:13
  Altman announced an agreement with the Department of Defense
• 18:16
  to deploy their models on classified networks, explicitly
• 18:19
  agreeing to the military's terms without those specific ethical
• 18:22
  restrictions regarding autonomous targeting.
• 18:25
  Anthropic responded with massive legal retaliation, though.
• 18:27
  Oh. Yeah, the lawsuits.
• 18:29
  They filed lawsuits in both California and the DC Circuit.
• 18:33
  In the California case, Judge Rita F Flynn granted A
• 18:37
  preliminary injunction against the government.
• 18:39
  OK, She ruled that the government's constituted classic
• 18:44
  illegal First Amendment retaliation.
• 18:46
  Wow. Yeah, noting that internal
• 18:49
  government records revealed the supply chain risk designation
• 18:52
  was based largely on the company interacting in a hostile manner
• 18:55
  through the press, rather than any actual technological
• 18:59
  vulnerability. That's fascinating.
• 19:00
  But in the DC Circuit, the court refused to lift the designation.
• 19:05
  They prioritized the military's immediate operational needs
• 19:08
  during an ongoing conflict, specifically citing the
• 19:11
  requirements of Operation Epic Fury.
• 19:13
  I have to say I questioned the internal coherence of Anthropics
• 19:17
  ethical stance in this situation.
• 19:19
  They allowed their technology to be utilized for complex missile
• 19:22
  defense, logistics and high level cyber warfare planning,
• 19:25
  but they draw a hard, uncompromising line at
• 19:28
  autonomous targeting. When the technology is already
• 19:31
  integrated that deeply into the kill chain of military
• 19:34
  operations, Drawing an arbitrary line at the final trigger pole
• 19:38
  seems highly contradictory. I see your point, but a private
• 19:42
  enterprise has the right and honestly, the fundamental
• 19:45
  obligation to enforce its own terms of service.
• 19:49
  Even against the government. They built the underlying
• 19:51
  infrastructure, and they should be able to dictate the absolute
• 19:55
  boundaries of its use, even when negotiating against a global
• 19:58
  superpower. If the creators believe fully
• 20:01
  autonomous targeting crosses an unbreachable moral line, they
• 20:05
  must enforce that boundary regardless of the financial
• 20:08
  cost. Well, this completely rewrites
• 20:11
  the rules of engagement between Silicon Valley and the Pentagon.
• 20:14
  It really does. It opens up incredibly lucrative
• 20:17
  long term defense contracts for artificial intelligence
• 20:20
  companies is willing to waive their ethical restrictions while
• 20:23
  simultaneously punishing those who attempt to dictate
• 20:26
  operational terms to the military.
• 20:27
  Apparatus. We are seeing a rapidly
• 20:29
  developing resolution to the standalk, however.
• 20:32
  Oh, the White House pivot. Yeah, according to industry
• 20:35
  sources, the White House is currently drafting plans to
• 20:38
  completely bypass the supply chain Risk designation,
• 20:42
  specifically to permit federal use of the Mythos AI model we
• 20:45
  discussed earlier. The reasoning behind this pivot
• 20:48
  is purely pragmatic. Totally Government cybersecurity
• 20:51
  agencies and intelligence branches quickly recognize that
• 20:54
  they simply could not afford to be locked out of the most
• 20:56
  advanced defensive cyber tool available on the market.
• 21:00
  Exactly. When your adversaries are
• 21:02
  rapidly deploying similar autonomous capabilities,
• 21:05
  handicapping your own defensive infrastructure over a standard
• 21:07
  contract dispute is a completely untenable strategic position.
• 21:12
  This severely limits the government's ability to
• 21:14
  successfully blacklist critical technology providers.
• 21:18
  It really does. When the state's own defensive
• 21:20
  capabilities rely entirely on that private infrastructure to
• 21:23
  function, the threat of a ban loses all leverage.
• 21:27
  You cannot effectively ban the company that holds the digital
• 21:30
  keys to your own national security.
• 21:32
  The trajectory of Anthropic proves the real value of
• 21:35
  artificial intelligence lies directly in enterprise
• 21:39
  infrastructure and automated coding, creating an entity so
• 21:43
  vital to the global economy that it forces governments to rewrite
• 21:47
  their own rules of procurement. If a private company controls
• 21:50
  the automated intelligence that patches our critical
• 21:53
  infrastructure, writes our corporate software, and
• 21:56
  negotiates terms of engagement with the military, who really
• 21:59
  holds the ultimate authority in the modern world, the government
• 22:03
  or the algorithm? If you're not subscribed yet,
• 22:05
  take a second and hit follow on whatever app you're using.
• 22:08
  It helps us keep making this. We appreciate you being here.
• 22:10
  Also check out our YouTube channel for more business and
• 22:13
  tech updates. There's a link in the
• 22:14
  description.