Latest / Elon Musk Podcast / Anthropic beats Pentagon ban for $60B IPO

Transcript

Anthropic beats Pentagon ban for $60B IPO · Elon Musk Podcast

• 0:00
  Anthropic just secured a federal injunction that stops the Trump
• 0:03
  administration and the Pentagon from labeling the AI company a
• 0:07
  supply chain risk and a threat to US National Security.
• 0:11
  Right. Yeah, because President Donald
• 0:13
  Trump had issued this directive banning federal agencies from
• 0:16
  using Anthropic's clawed models across the board.
• 0:19
  So Anthropic's sued immediately, citing severe monetary and
• 0:23
  reputational harm, and a federal judge in San Francisco, Rita
• 0:26
  Lynn, granted A preliminary injunction to halt the
• 0:30
  government's actions. Which really sets up this wild
• 0:32
  scenario. How does a tech company balance
• 0:35
  fighting a massive legal battle with the military while
• 0:38
  simultaneously trying to beat its biggest wivals to a
• 0:41
  record-breaking stock market debut?
• 0:44
  Well, I mean, it requires an incredibly aggressive strategy
• 0:46
  on multiple fronts. It's here starting with
• 0:48
  basically neutralizing that immediate existential threat
• 0:51
  because Judge Rita Lynn didn't just, you know, pause the ban,
• 0:54
  she severely hampered the Pentagons efforts by ruling that
• 0:57
  the government actions were actually illegal.
• 0:59
  Hold on, wait, back up. How exactly does a government
• 1:02
  contract dispute become like a constitutional issue?
• 1:05
  Is this similar to a whistleblower facing retaliation
• 1:08
  for speaking out? Yes, exactly that.
• 1:11
  It's a great analogy, Judge Lynn actually wrote in her order that
• 1:14
  punishing Anthropic for bringing public scrutiny to the
• 1:17
  government's contracting position is, quote, classic
• 1:21
  illegal First Amendment retaliation.
• 1:24
  I'm still, I'm struggling with this First Amendment angle a
• 1:27
  little bit because I always thought the government could
• 1:29
  just choose who it does business with.
• 1:31
  Like, if I don't buy a certain brand of coffee, I'm not
• 1:34
  violating the barista's free speech, right?
• 1:36
  Yeah. Why is the Pentagon legally
• 1:38
  obligated to keep the door open for Anthropic?
• 1:40
  That's a totally fair question. The distinction lies in the
• 1:43
  mechanism of state power versus, you know, private preference.
• 1:47
  As a private citizen, you can boycott a coffee shop because
• 1:50
  you just dislike their espresso. The federal government, however,
• 1:53
  operates under the Constitution. Specifically, there's this thing
• 1:56
  called the unconstitutional conditions doctrine.
• 2:00
  They absolutely cannot leverage their enormous purchasing power
• 2:03
  to silence criticism or punish dissent.
• 2:06
  Yeah. And according to the court
• 2:08
  documents, Anthropic had been highly vocal.
• 2:11
  They were bringing public scrutiny to how the government
• 2:13
  awards its contracts and structures its tech
• 2:16
  acquisitions. So that refines the entire
• 2:18
  situation. The core argument is that the
• 2:21
  government was basically using its purchasing power to punish a
• 2:24
  private company for drawing attention to the inner workings
• 2:27
  of federal procurement. Precisely when a company
• 2:30
  challenges a federal contracting directive, they are engaging in
• 2:33
  protected speech. The court basically found that
• 2:35
  the administration's response labeling Anthropic a threat to
• 2:39
  US National Security and a supply chain risk wasn't a
• 2:42
  legitimate security measure based on like, software
• 2:46
  vulnerabilities. Right, they didn't find a bug in
• 2:48
  the code. Exactly.
• 2:49
  It was interpreted by the court as a retaliatory strike designed
• 2:53
  to silence that scrutiny. And a supply chain risk label
• 2:57
  from the Pentagon. I mean, that is not just a
• 2:59
  polite decline of services. It basically functions as a
• 3:02
  blacklist. Oh, it's a total freeze.
• 3:04
  Right, because if the Department of Defense says a piece of
• 3:06
  software is a national security threat, that designation
• 3:10
  radiates outward immediately. It tells every other federal
• 3:13
  agency like the Department of Energy, the Treasury, Homeland
• 3:16
  Security to just freeze their procurement.
• 3:19
  And it goes even further than domestic agencies, too, that
• 3:22
  label signals to allied nations across Europe and Asia to avoid
• 3:27
  integrating the technology into their own infrastructure.
• 3:30
  So it kills international contracts as well.
• 3:32
  Yeah, and it even makes private enterprise incredibly nervous.
• 3:36
  I mean, if you are a multinational bank or a massive
• 3:38
  healthcare provider, you absolutely do not want to build
• 3:41
  your internal systems on a foundation that the US military
• 3:45
  has formally designated as a risk.
• 3:46
  No, of course not. The compliance officers would
• 3:48
  have a field day with that. Right.
• 3:50
  So by securing this injunction based on 1st Amendment
• 3:52
  protections, Anthropic essentially neutralized a weapon
• 3:56
  that would have systematically dismantled their entire revenue
• 3:59
  stream. It really sets a fascinating
• 4:02
  precedent for how tech companies can push back against unilateral
• 4:06
  executive directives. Like they aren't just arguing a
• 4:10
  standard breach of contract or an administrative error, they
• 4:13
  are successfully arguing a violation of fundamental
• 4:16
  constitutional rights to protect their commercial viability.
• 4:20
  So the ban is paused and the government has to back off.
• 4:23
  Which is huge, because if the government can unilaterally
• 4:26
  blacklist your product, Wall Street isn't going to give you a
• 4:29
  $60 billion valuation, no. Not a chance.
• 4:32
  That First Amendment victory wasn't just about constitutional
• 4:34
  rights. It was the strict prerequisite
• 4:37
  for what Anthropic is doing next.
• 4:39
  Because despite all these intense legal battles, Anthropic
• 4:42
  is actively preparing for a massive IPO.
• 4:46
  They're lining up bankers, securing legal counsel, and
• 4:49
  they're pitching a public offering that could raise over
• 4:51
  60 billion dollars. Yeah. $60 billion that that
• 4:56
  alters the financial calculus of the entire technology sector.
• 4:59
  That is an astronomical amount of money it.
• 5:01
  Really is an initial public offering of that magnitude
• 5:05
  requires an unprecedented level of institutional coordination.
• 5:09
  We are talking about marshalling the largest investment banks in
• 5:12
  the world to convince mutual funds, state pension funds,
• 5:16
  sovereign wealth portfolios to allocate massive, massive
• 5:21
  amounts of capital into a single, highly specialized
• 5:24
  artificial intelligence entity. And Anthropic isn't the only one
• 5:28
  eyeing a mega IPO right now. No, not at all.
• 5:30
  You've got Challenger, SpaceX and archrival Open AI making
• 5:34
  very similar moves. Yeah, the capital markets are
• 5:36
  suddenly looking at multiple generational companies all
• 5:39
  trying to access public liquidity at the exact same
• 5:41
  moment. And there's a really specific
• 5:44
  pick behind the scenes detail here from the sources Open
• 5:46
  AICEO. Sam Altman has privately stated
• 5:50
  that he wants Open AI to list its shares on the stock market
• 5:53
  before Anthropic. Does, which tells you everything
• 5:55
  you need to know about the stakes here.
• 5:57
  The pressure to go public isn't just about raising capital
• 6:00
  anymore. It is a direct race to capture
• 6:02
  public investor money before a major rival camp.
• 6:05
  It's a land grab. Exactly.
• 6:07
  This involves the strict mechanics of how institutional
• 6:10
  capital is deployed. Like if you are listening to
• 6:12
  this and you have a four O 1K or a pension plan, the managers of
• 6:16
  those funds operate on mathematically defined risk
• 6:19
  mandates. I kind of think about it like 2
• 6:22
  massive summer blockbusters trying to claim the exact same
• 6:26
  opening weekend. That's a good way to look.
• 6:28
  At it, there is only so much attention and more importantly,
• 6:31
  there is only so much liquid capital available at any given
• 6:34
  moment for high risk, high reward tech investments.
• 6:37
  Right, because institutional investors have strict allocation
• 6:40
  limits. A major pension fund.
• 6:43
  They might decide they want exactly 3% of their total
• 6:46
  portfolio exposed to foundational AI models.
• 6:49
  OK, they are not going to buy every single AI company that
• 6:52
  goes public if open AI gets out the door first.
• 6:55
  They establish the baseline and they absorb the vast majority of
• 6:58
  that allocated capital. The fund manager fills their 3%
• 7:01
  quota and the vault is essentially closed.
• 7:03
  Wow, so if Open AI lists fork, they literally set the terms for
• 7:07
  the entire industry. If they command a certain
• 7:10
  multiple on their revenue, say 50 times their projected
• 7:14
  earnings, Anthropic is then forced to either mathematically
• 7:18
  justify to those same fund managers why they deserve a
• 7:21
  higher multiple, or they just have to accept a lower valuation
• 7:25
  based on the benchmark Open AI just established.
• 7:28
  It completely anchors the market.
• 7:30
  Being second means you are constantly compared to the first
• 7:33
  mover, rather than being evaluated purely on your own
• 7:36
  isolated financial merits. Which puts Anthropic in a really
• 7:39
  tough spot. Yeah, that private statement
• 7:41
  from Altman really highlights that the competition between
• 7:44
  these AI labs isn't just about who builds the smartest model
• 7:47
  anymore. It is fundamentally about who
• 7:49
  controls the narrative on Wall Street.
• 7:51
  You know, hearing all this, you would naturally assume that
• 7:54
  beating the government in federal court over a First
• 7:56
  Amendment retaliation claim was the hardest hurdle for
• 7:58
  Anthropics IPOI. Actually disagree completely.
• 8:01
  Really. Yeah, I mean, defeating a
• 8:03
  blacklist is obviously a massive legal victory, but the true
• 8:06
  hurdle is going to be the financial scrutiny from
• 8:08
  regulators, not the Pentagon. Interesting.
• 8:11
  Why do you say that? Well, a judge assesses legal
• 8:14
  arguments, right? The Securities and Exchange
• 8:17
  Commission assesses pure financial reality and up until
• 8:21
  now these companies have been incredibly secretive about their
• 8:25
  internal economics. Because they're private.
• 8:27
  Right, They operate as private entities backed by venture
• 8:30
  capital, where growth is basically the only metric that
• 8:34
  drives private funding. But to go public, they have to
• 8:37
  open the books. Exactly.
• 8:39
  And regulators are going to closely examine the differing
• 8:41
  methods Open AI and Anthropic use to actually book their
• 8:45
  revenue, specifically revenue related to cloud providers that
• 8:49
  resell their AI models. OK, let's look at how this money
• 8:52
  actually flows, because the accounting gets incredibly
• 8:55
  complex when third parties are involved.
• 8:57
  Oh, it's a nightmare. It requires looking at the
• 8:59
  mechanics of gross versus net revenue reporting.
• 9:02
  Right, so say I build a software application for logistics and I
• 9:05
  want to integrate an AI model to process all my data.
• 9:08
  I might not go directly to Anthropic, I might use a major
• 9:11
  cloud provider like Amazon Web Services or Google Cloud because
• 9:14
  my data is already hosted on their servers anyway.
• 9:17
  Yeah. So you pay the cloud provider
• 9:19
  for access to the AI. The cloud provider then owes a
• 9:22
  portion of that money to the company that actually built the
• 9:25
  model. So in for optic.
• 9:27
  Right. The critical question for the
• 9:28
  SEC is how the AI company records that specific
• 9:32
  transaction on their income statement.
• 9:35
  Do they record the total amount the end user paid as top line
• 9:39
  revenue and then list the cloud providers cut as a separate
• 9:43
  expense? Or do they only record the net
• 9:46
  amount that actually hits their bank account?
• 9:48
  It is kind of like selling a house.
• 9:50
  If you sell a property for $500,000 but your real estate
• 9:54
  agent takes a $30,000 Commission cut, do you tell your bank you
• 9:57
  made half a million or 470,000? That's a perfect example.
• 10:01
  And if an AI company records the gross amount, their top line
• 10:04
  revenue looks artificially inflated.
• 10:06
  Because it creates the illusion that their direct sales volume
• 10:09
  and market dominance are much higher than they actually are.
• 10:11
  Exactly. And in the lead up to an IPO,
• 10:14
  top line revenue growth is the primary metric investment
• 10:17
  bankers use to justify these massive $60 billion valuation,
• 10:21
  of course. So regulators are tasked with
• 10:23
  ensuring the public investors are getting an accurate,
• 10:26
  transparent view of a company's financial health.
• 10:30
  If differing AI companies are using completely different
• 10:34
  accounting standards for their cloud partnerships, it becomes
• 10:37
  mathematically impossible for the public market to accurately
• 10:40
  compare them. Wait.
• 10:41
  Let's look at the infrastructure side of this too, because these
• 10:44
  companies also have to justify how they are projecting
• 10:47
  unprecedented revenue growth while managing massive server
• 10:51
  costs. You cannot separate the software
• 10:54
  revenue from the underlying physical infrastructure required
• 10:57
  to generate it, and the cloud providers reselling the AI
• 11:00
  models are often the exact same companies providing the servers.
• 11:03
  This is where it gets really murky.
• 11:05
  The server costs are the hidden anchor on these balance sheets.
• 11:08
  Training a frontier AI model requires billions of dollars in
• 11:12
  raw compute power, right? And then running that model for
• 11:15
  millions of users every day requires ongoing immense
• 11:19
  expenditures in electricity, cooling, hardware, maintenance.
• 11:21
  So Anthropic pays the cloud provider massive sums for
• 11:25
  computing power to train the model, and then the cloud
• 11:28
  provider pays Anthropic for access to sell that same model
• 11:31
  to its users. Yes.
• 11:33
  That's an Infinity loop of money.
• 11:35
  If I am an SEC regulator, how do I know any of this is real
• 11:39
  profit and not just two tech giants trading the same $100
• 11:42
  bill back and forth? You hit the nail on the head.
• 11:45
  Exposing this internal math is exactly what will invite
• 11:48
  unprecedented scrutiny. To go public, these notoriously
• 11:52
  private AI companies will have to prove to the SEC that their
• 11:55
  valuation is based on solid financial reality rather than
• 11:59
  circular accounting. They have to prove the cash is
• 12:02
  real. Exactly.
• 12:03
  Regulators will demand to see if these partnerships are
• 12:05
  structurally sound or if they rely on circular revenue flows,
• 12:09
  because sometimes they are even trading compute credits instead
• 12:11
  of actual cash, just to mask the true cost of operating
• 12:15
  artificial intelligence at scale.
• 12:16
  So to sum this all up, Anthropic is basically fighting a multi
• 12:19
  front war. They are fending off a
• 12:21
  government blacklist with a First Amendment defense while
• 12:24
  simultaneously racing open AI to a massive public offering.
• 12:27
  And it leaves you wondering, you know, when securities regulators
• 12:30
  finally force these AI giants to reveal exactly how they count
• 12:33
  their cloud revenue and server costs, will that $60 billion
• 12:37
  price tag actually hold up? If you're not subscribed yet,
• 12:40
  take a second and hit follow on whatever app you're using.
• 12:43
  It helps us keep making this. We appreciate you being here.