Latest / Elon Musk Podcast / Oracle Fired 30,000 to Build AI

Transcript

Oracle Fired 30,000 to Build AI · Elon Musk Podcast

• 0:00
  Oracle eliminated up to 30,000 jobs globally through a single
• 0:05
  automated early morning e-mail to fund 1/2 a trillion dollar
• 0:10
  artificial intelligence project with Open AI and SoftBank.
• 0:14
  Yeah, that is roughly 18% of 162,000 person workforce just
• 0:20
  wiped out. I mean, before most of them had
• 0:21
  even had their morning coffee, right?
• 0:23
  It's just massive. It really is.
• 0:25
  And looking through the stack of internal memos, the financial
• 0:28
  disclosures, and, you know, the geopolitical analysis, today we
• 0:33
  are really focusing on the raw mechanics of this.
• 0:35
  We are looking at the crushing financial pressures of building
• 0:38
  artificial intelligence infrastructure and how this is
• 0:41
  sending shockwaves straight through the technology and
• 0:44
  healthcare sectors. Yeah.
• 0:45
  And the question we're going to be trying to answer through all
• 0:47
  of this is how does a company justify hollowing out its own
• 0:50
  human workforce to build the machines meant to replace them?
• 0:54
  Right. Well, that clinical execution of
• 0:55
  these layoffs is what really illustrates the underlying
• 0:57
  strategy here. We are not talking about
• 0:59
  standard corporate restructuring where, you know, human resources
• 1:03
  schedules a calendar invite. Right.
• 1:05
  Or will you sit down and discuss severance package?
• 1:07
  Exactly. These employees received an
• 1:09
  e-mail signed simply from Oracle leadership before gone System
• 1:13
  access was revoked instantly. Yeah.
• 1:16
  There were no meetings. No periods, you literally just
• 1:19
  wake up and your job is gone. And looking through the sources,
• 1:23
  the internal Reddit and blind forums were just completely
• 1:26
  chaotic. Oh, absolutely.
• 1:28
  For anyone unfamiliar, Blind is an anonymous platform where
• 1:31
  verified employees internal company matters and it just lit
• 1:35
  up with panic. We are talking about reports of
• 1:38
  complete teens being absolutely decimated.
• 1:40
  Yeah, entire departments. Right.
• 1:42
  And you had directors allegedly feigning ignorance, like acting
• 1:46
  like they had no idea this was coming, while their entire
• 1:48
  departments vanished overnight. That creates an environment of
• 1:51
  total paranoia. It was a highly coordinated
• 1:53
  operation, clearly, but what truly terrified the workforce
• 1:57
  was the discovery of an internal spyware tool.
• 2:01
  It was called IT Client Utility. Yeah, I read this part of the
• 2:04
  notes and I had to read it twice to make sure I was understanding
• 2:07
  the mechanics of it correctly. I don't blame you.
• 2:10
  Basically, before the termination emails went out,
• 2:12
  this software was autoconfigured to change employee privacy
• 2:16
  settings. It secifically allowed screen
• 2:19
  and audio recording. Wait, back U they installed
• 2:21
  software to record screens right before cutting off access
• 2:25
  entirely. Exactly.
• 2:27
  They used an automated deployment to silently force
• 2:30
  those privacy permissions open. That is insane.
• 2:33
  Right. In any normal operating system,
• 2:35
  if a program wants to use your microphone or record your
• 2:37
  screen, you get a pop up, right? You get asked for permission.
• 2:40
  This utility bypassed that entirely.
• 2:43
  Jeez. It effectively turned employee
• 2:45
  hardware into surveillance devices in the final hours of
• 2:48
  their employment. The corporate logic behind it is
• 2:51
  to guarantee zero data exfiltration.
• 2:54
  Right, they want to ensure no disgruntled employee downloads
• 2:56
  proprietary source code or customer databases on their way
• 3:00
  out the door. Exactly, but the method is
• 3:02
  incredibly invasive. I mean, imagine going to sleep
• 3:04
  thinking you have a job and while you were dreaming your
• 3:07
  company remotely alters your webcam and microphone
• 3:10
  permissions just to fire you. It's dystopian.
• 3:13
  It is a completely chilling level of corporate paranoia.
• 3:17
  It is treating your employees the exact same way you treat
• 3:21
  legacy software code being abruptly deprecated.
• 3:24
  You just switch it off, wipe the server and move on.
• 3:26
  Yeah, perfectly stated. There is absolutely no human
• 3:30
  element left in that equation. You do not treat people like
• 3:33
  that unless you view them purely as an expense on a spreadsheet.
• 3:37
  And handling terminations this way.
• 3:40
  It destroys internal trust completely.
• 3:42
  It forces the remaining staff into a bare bones survival mode.
• 3:46
  It permanently alters the employer employee relationship
• 3:49
  into a strictly transactional contract.
• 3:51
  Nobody is going above and beyond after seeing that.
• 3:53
  No, exactly. The people left behind are not
• 3:56
  going to be innovating or taking creative risks.
• 3:58
  They're just going to be looking over their shoulders.
• 4:00
  Well, we have to look at the severe financial realities
• 4:02
  driving a workforce reduction of this magnitude.
• 4:05
  Oracle needs to free up roughly eight $10 billion in cash flow.
• 4:09
  Yeah, the math here is just punishing.
• 4:11
  The company is carrying a debt load exceeding $100 billion.
• 4:15
  Hold on $100 billion in debt. Yes, $100 billion, and they're
• 4:22
  also facing a negative free cash flow of over $13 billion.
• 4:26
  Wow. And at the exact same time, they
• 4:28
  just recorded nearly a billion dollars in a restructuring plan.
• 4:32
  For you listening to this, when we talk about negative free cash
• 4:36
  flow of $13 billion, we're not just saying profits are down.
• 4:40
  Right, it's worse than that. Yeah, We are saying the money
• 4:42
  coming in is severely short of the money going out to maintain
• 4:46
  operations and investments. They're actively burning through
• 4:50
  reserves just to keep the lights on and build these new data
• 4:53
  centers. Precisely.
• 4:54
  And that $100 billion in debt, it requires constant servicing,
• 5:00
  right? You have to pay the interest
• 5:01
  you. Have to pay the interest on that
• 5:02
  debt regardless of how your software sales are doing.
• 5:05
  When a company is carrying that much leverage while
• 5:07
  simultaneously bleeding $13 billion in negative free cash
• 5:11
  flow, traditional operational spending becomes a luxury they
• 5:14
  simply cannot afford. So the payroll of 30,000
• 5:17
  employees is the fastest way to stop the bleeding and reroute
• 5:20
  capital. So they are torching through
• 5:22
  cash to finance artificial intelligence data centers, and
• 5:26
  those human salaries are being directly converted into compute
• 5:30
  infrastructure. Instead of paying a software
• 5:32
  engineer in California, they're using that same money to buy
• 5:36
  server racks, land and cooling vans.
• 5:38
  It is a massive capital expenditure shift.
• 5:41
  Running a traditional enterprise software company requires, you
• 5:44
  know, human sales teams, human support staff, human developers.
• 5:48
  Yeah. Running a cloud computing
• 5:50
  utility just requires electricity and graphics
• 5:53
  processing units, Right. This extreme financial strain
• 5:56
  limits Oracle's ability to maintain its traditional
• 5:59
  enterprise software operations. It forces an all or nothing
• 6:03
  financial commitment onto raw compute power and
• 6:05
  infrastructure. They are betting the entire farm
• 6:07
  on this One Direction. They absolutely are.
• 6:09
  And the direction is Stargate Oracle is part of Stargate LLCA
• 6:13
  $500 billion joint venture with Open AI and SoftBank. $500
• 6:19
  billion is half a trillion dollars, and the players
• 6:22
  involved here are just fascinating.
• 6:23
  President Donald Trump announced this initiative alongside Sam
• 6:26
  Altman, Larry Ellison, and Masayoshi Sun.
• 6:29
  Yeah, the entire project is being framed as a national
• 6:32
  security imperative to secure American technology leadership.
• 6:36
  Exactly. By grouping the White House, the
• 6:39
  algorithm creators at Open AI, the server infrastructure of
• 6:43
  Oracle, and the capital of SoftBank, they are presenting
• 6:46
  this not just as a business deal, but as an essential pillar
• 6:50
  of the country's defense and technological dominance.
• 6:53
  Right. And Larry Ellison claims this
• 6:54
  computing power could lead to robotically designed mRNA cancer
• 6:59
  vaccines in a matter of days. That is the kind of promise they
• 7:02
  are using to justify the investment.
• 7:04
  We really need to explain how artificial intelligence actually
• 7:07
  does that though. Yeah, go.
• 7:08
  Ahead An mRNA vaccine relies on folding proteins in very
• 7:12
  specific, complex ways. Figuring out the right structure
• 7:15
  manually or with traditional computers can literally take
• 7:19
  years. Because there's so many
• 7:20
  variations. Exactly.
• 7:21
  With the kind of raw compute power Stargate promises,
• 7:25
  algorithms can simulate millions of protein folds per second.
• 7:29
  So they are effectively testing and designing vaccines digitally
• 7:32
  before they ever hit a physical lab.
• 7:34
  It is basically like the construction of the Interstate
• 7:37
  highway system, but for digital processing power.
• 7:41
  You are laying down the physical concrete and steel required for
• 7:44
  the next century of commerce and medicine.
• 7:47
  That is a great way to look at it, but you also have to view
• 7:50
  this against the backdrop of an ongoing US Iran war that is
• 7:54
  disrupting global fuel and tech markets.
• 7:57
  Right, because building domestic data centers of this magnitude
• 8:01
  requires enormous amounts of energy.
• 8:03
  You cannot run a half a trillion dollar supercomputer on a few
• 8:07
  solar panels. Not at all.
• 8:09
  These facilities consume power on the scale of entire cities.
• 8:12
  That makes the geopolitical stability of fuel supplies a
• 8:15
  critical factor in whether these facilities can actually operate.
• 8:18
  The energy grid is everything. Yeah, the conflict directly
• 8:21
  impacts the global energy grid and the supply chains needed to
• 8:24
  build and power these data centers.
• 8:26
  This massive build out opens up vast national security and
• 8:29
  medical possibilities, but it fundamentally ties corporate
• 8:32
  success directly to geopolitical stability.
• 8:35
  If the energy grid stumbles because of international
• 8:37
  conflict, the entire half a trillion dollar machine grinds
• 8:40
  to a halt. Precisely.
• 8:42
  And we are actually already seeing friction threatening the
• 8:45
  partnership right now. Despite the massive financial
• 8:48
  commitment, the Stargate rollout is stalling.
• 8:51
  Oracle, Open AI and Soft A are squabbling over who gets
• 8:54
  ultimate control of the facilities.
• 8:56
  Yeah, when you have 3 entities contributing massive distinct
• 8:59
  assets, determining the hierarchy of control becomes
• 9:03
  incredibly complicated. Oracle has the land and the
• 9:06
  servers open. AI has the algorithms.
• 9:10
  SoftBank has the capital. And you have founders and
• 9:12
  executives with historically massive egos trying to share the
• 9:16
  steering wheel on the most expensive infrastructure project
• 9:19
  in tech history. Oh, absolutely.
• 9:21
  Because of this infighting, a massive 600 MW expansion in
• 9:24
  Abilene, TX was scrapped entirely.
• 9:27
  Yeah, for context on that number, 600 megawatts is enough
• 9:30
  electricity to power hundreds of thousands of homes.
• 9:33
  It is a staggering amount of infrastructure.
• 9:36
  And because they scrapped it, that land an opening for rival
• 9:39
  Meta to step in and lease the site.
• 9:41
  Which is just a huge loss for them.
• 9:43
  I actually think the partnership will stabilize, though.
• 9:46
  They have too much capital committed to let it fail.
• 9:49
  And these are just the growing pains of a massive joint
• 9:52
  venture. You get Laurie Ellison, Sam
• 9:54
  Altman and Masayoshi Son in the same room.
• 9:57
  You were going to have some arguments over who sits at the
• 9:59
  head of the table, but they will figure it out.
• 10:02
  See, I disagree completely. Yeah, three massive entities
• 10:06
  sharing 1/2 a trillion dollar pie was inevitably going to
• 10:10
  fracture over control. The Abilene cancellation proves
• 10:14
  they cannot execute smoothly. Well, there are always bumps in
• 10:18
  the road, right? Sure, but when you have an
• 10:20
  investment this large, speed is everything.
• 10:22
  If you are arguing over facility control while your competitors
• 10:25
  are building, you lose. Fairpoint.
• 10:27
  Either way, losing a 600 MW site to Meta is a brutal unforced
• 10:32
  error. You do not just find another
• 10:34
  power grid of that size waiting around to be leased.
• 10:36
  No, you don't. This internal conflict slows
• 10:39
  down the deployment of essential infrastructure.
• 10:41
  It opens up immediate opportunities for direct
• 10:43
  competitors to cease physical data center space while the
• 10:46
  Stargate partners are busy arguing in the boardroom.
• 10:49
  Right, because Meta did not have to negotiate a complex three-way
• 10:53
  joint venture. Exactly, they just walked in and
• 10:55
  signed the lease. And while all this new
• 10:58
  infrastructure is failing to launch smoothly, the older
• 11:01
  legacy divisions at Oracle are completely collapsing.
• 11:05
  A few years ago, Oracle spent $28 billion to acquire the
• 11:09
  healthcare technology firm Cerner.
• 11:12
  Yeah, that acquisition was supposed to transform hospital
• 11:15
  record keeping. The goal was to take Cerner's
• 11:18
  massive footprint in the medical field and modernize it.
• 11:22
  But that hasn't happened. No, now they have lost 57 acute
• 11:26
  care customers, which includes 12 massive hospital networks.
• 11:30
  And just so we're clear, acute care is the intense immediate
• 11:33
  treatment you get in an emergency room or a surgical
• 11:36
  ward. The communication and the actual
• 11:38
  software partnerships there have just deteriorated.
• 11:40
  Completely and five senior cloud executives who were specifically
• 11:44
  transferred to modernized Cerner's aging systems have just
• 11:47
  left the company. Well, they spent $28 billion to
• 11:50
  transform healthcare, and now the executives sent to fix it
• 11:53
  are walking out. They are exiting because the
• 11:55
  strategy fundamentally clashed with the reality of hospital
• 11:58
  operations. Oracle completely rewrote Cerner
• 12:03
  software code to integrate it into their broad cloud
• 12:05
  architecture. OK, that sounds like a normal
• 12:07
  tech upgrade though. Right, but acute care workflows
• 12:10
  are highly specific. A hospital cannot operate like a
• 12:13
  standard corporate logistics firm.
• 12:16
  It integrates with heart monitors, 4 pumps and lab
• 12:19
  results in real time. If a cloud server has latency, a
• 12:23
  doctor does not get the alert. Wow, so it is a life or death
• 12:26
  issue? Exactly.
• 12:28
  When you try to force complex life or death medical software
• 12:31
  onto a generic cloud platform, things break.
• 12:34
  And because things are breaking, the customers are fleeing.
• 12:36
  The resources required to fix a failing, highly specialized
• 12:40
  medical software company are enormous, but those resources
• 12:43
  are now being cannibalized to buy graphics processing units
• 12:46
  for Stargate. Yes, exactly.
• 12:49
  It really shows you what happens when a company decides its pass.
• 12:51
  Acquisitions are no longer the priority.
• 12:53
  They're basically stripping the copper out of the walls of their
• 12:56
  healthcare division to pay for the artificial intelligence
• 12:58
  project. And this failure fundamentally
• 13:01
  changes Oracle's corporate identity.
• 13:04
  It marks the definitive end of their attempt to dominate
• 13:06
  hospital software and completes their shift toward being a raw
• 13:10
  cloud utility provider. Right.
• 13:12
  They are no longer interested in building the software you use,
• 13:15
  they just want to rent you the servers that run it.
• 13:18
  Which brings us to what this means for the people actually
• 13:21
  doing the work. Yeah, Chief information officers
• 13:24
  are now forced to run IT ecosystems with significantly
• 13:28
  fewer people. Oh absolutely.
• 13:29
  They are having to ruthlessly prioritize mission critical
• 13:33
  operations like security and incident response.
• 13:35
  Yeah, the nice to have projects are dead.
• 13:37
  Completely dead. There are no more internal
• 13:39
  portals or experimental software rollouts.
• 13:42
  It is entirely about keeping the network secure and the servers
• 13:45
  running with a fraction of the historical headcount.
• 13:48
  Meanwhile, displaced tech workers are realizing that to
• 13:51
  survive in this job market, they must adapt by securing specific
• 13:55
  cloud certifications from AWS or Google Cloud.
• 13:58
  And the irony here is incredible.
• 14:00
  These workers were just fired because their company decided to
• 14:03
  pivot to artificial intelligence infrastructure, and now they
• 14:07
  must learn artificial intelligence skills just to
• 14:09
  survive the layoffs caused by artificial intelligence.
• 14:13
  It is the new baseline, and for you listening to this, consider
• 14:16
  how relying on heavily automated systems and vibe coding makes
• 14:20
  your organization vulnerable if a single artificial intelligence
• 14:24
  vendor fails. We should probably explain vibe
• 14:27
  coding. Yeah, please do.
• 14:28
  This is the practice of using artificial intelligence to
• 14:31
  generate massive amounts of software code simply by feeding
• 14:35
  the machine natural language prompts.
• 14:37
  You just type what you want and the machine writes it right.
• 14:40
  The problem is human engineers are not writing the
• 14:43
  documentation or fully understanding the underlying
• 14:45
  architecture. Exactly.
• 14:47
  So if your entire operational stack relies on code generated
• 14:51
  by a machine, and that vendor experiences an outage or goes
• 14:54
  bankrupt, you no longer have the human engineers required to fix
• 14:58
  the underlying architecture. Because they don't know how it
• 15:00
  was built. Right.
• 15:01
  The people who understood how to manually write and repair that
• 15:05
  card have all been laid off. You are entirely at the mercy of
• 15:08
  the infrastructure provider if the artificial intelligence goes
• 15:11
  down. You cannot just roll up your
• 15:14
  sleeves and fix it yourself because you never wrote it in
• 15:17
  the first place. And this severely limits
• 15:19
  traditional career paths in tech.
• 15:22
  It forces professionals to become hyper specialized
• 15:25
  managers of automated systems rather than traditional
• 15:28
  creators. You are no longer building the
• 15:30
  house, you are just monitoring the thermostat.
• 15:34
  A traditional software giant just burned down its own
• 15:36
  workforce and pass acquisitions to buy a seat at the
• 15:40
  infrastructure table. They effectively traded human
• 15:43
  institutional knowledge for raw computing power.
• 15:46
  Will the eventual profits from renting out this computing power
• 15:49
  actually be enough to cover the massive debt and the permanent
• 15:53
  loss of human capital? If you're not subscribed yet,
• 15:55
  take a second and hit follow on whatever app you're using.
• 15:58
  It helps us keep making this. We appreciate you being here.