Latest / Tech Talks With Kinsoft / Last Week in Tech - The Biggest Patch Tuesday Ever, a 10.0 With No Patch to Apply, and Oracle's $664 Billion Round Trip

Transcript

Last Week in Tech - The Biggest Patch Tuesday Ever, a 10.0 With No Patch to Apply, and Oracle's $664 Billion Round Trip · Tech Talks With Kinsoft

• 0:00
  Imagine a burglar walks up to your house, right?
• 0:03
  They don't pick the lock. They don't throw a
• 0:06
  brick through a window. They just wait. Exactly.
• 0:08
  They just wait for your smart doorbell to automatically
• 0:10
  send you a notification. And the mere act of
• 0:13
  your doorbell generating that alert causes your
• 0:15
  front door to just fly wide open. You never even
• 0:18
  pressed a button. It's a really visceral image.
• 0:22
  But it honestly perfectly captures the reality
• 0:25
  of the digital landscape right now. The very
• 0:28
  defenses we invest in. are becoming the doorways.
• 0:31
  And we're going to get into the specific mechanics
• 0:34
  of exactly how that is happening in today's deep
• 0:36
  dive. We're unpacking a single just densely packed
• 0:40
  source today. It's a mid -September 2026 industry
• 0:43
  update from Kinsoft called Last Week in Tech.
• 0:46
  Right. And our mission here is to extract the
• 0:48
  critical, you know, the need to knows for anyone
• 0:50
  who relies on digital infrastructure, which,
• 0:52
  let's be honest, that is every single one of
• 0:55
  us. Yeah. We are looking at a tech world that
• 0:58
  is currently buckling under extreme stress. I
• 1:00
  mean, from attackers finding mind -bendingly
• 1:04
  clever ways to slip past our perimeters to the
• 1:07
  sheer crushing weight of the money being poured
• 1:10
  into artificial intelligence right now. And if
• 1:12
  you are listening to this on your commute, maybe
• 1:14
  you're an IT director or maybe you run an e -commerce
• 1:16
  brand or you're just trying to understand where
• 1:18
  the Internet is headed. These hidden mechanics,
• 1:21
  they dictate your daily friction points. Absolutely.
• 1:23
  These aren't abstract concepts. These vulnerabilities,
• 1:26
  these financial shifts, they're the reason you
• 1:29
  might have to, I don't know, authorize weekend
• 1:31
  overtime for your team or even rewrite your entire
• 1:33
  hardware budget mid -quarter. Okay, let's unpack
• 1:36
  this. Starting with that doorbell analogy and
• 1:39
  how it applies to online retail. an online store
• 1:42
  on Magento or Adobe Commerce. The source material
• 1:45
  essentially issues a red alert. A massive red
• 1:48
  alert, yeah. In early September, attackers started
• 1:50
  exploiting a zero -day vulnerability dubbed Style
• 1:53
  Smuggler. And I want to make sure I understand
• 1:55
  the sheer severity of this because it has a CVSS
• 1:58
  score of 10 .0. That's the common vulnerability
• 2:00
  scoring system. And 10 is the absolute maximum,
• 2:04
  meaning it requires zero authentication and an
• 2:07
  attacker can execute code remotely. But what
• 2:10
  I don't get is how does it trigger without anyone
• 2:12
  actually clicking a malicious link? The mechanism
• 2:15
  is genuinely brilliant in a highly malicious
• 2:17
  way. Typically, an attacker tries to force malicious
• 2:21
  code through an input field, like a login box.
• 2:24
  Like typing code into a password field. Exactly.
• 2:27
  But here. They completely bypass that. They poison
• 2:30
  a standard log file that Magento naturally writes
• 2:33
  to just to keep track of its own activity. And
• 2:36
  they plant PHP code inside of it. Then they just
• 2:39
  trigger a completely mundane automated function.
• 2:42
  Specifically, it's a payment transaction failed
• 2:44
  reminder email. Wait, so let me make sure I have
• 2:46
  this straight. The email doesn't even need to
• 2:48
  be opened by an employee? Nope. Just the act
• 2:50
  of the systems thinking about rendering that
• 2:52
  normal message triggers the hack. Yes, exactly.
• 2:55
  When Magento simply attempts to assemble that
• 2:57
  email, it pulls from that poisoned log file and
• 3:01
  the malicious code executes. It's zero click
• 3:03
  in the purest sense. Nobody clicks a link. Nobody
• 3:06
  opens an attachment. Wow. And what drops into
• 3:08
  your system after that automated door opens is
• 3:11
  a highly sophisticated 1 .9 megabyte payload.
• 3:16
  And it's written in Rust. Which is super fast,
• 3:19
  right? Yeah. Rust is a programming language known
• 3:21
  for being incredibly fast and notoriously hard
• 3:24
  to reverse engineer. This program disguises itself
• 3:28
  as a normal Linux background task. It hides in
• 3:30
  a hidden directory above what's called the web
• 3:32
  root. Meaning what exactly? Basically, it stores
• 3:35
  itself outside the folder where your actual website
• 3:37
  files live. So your typical security scans. They
• 3:40
  completely miss it. Oh, wow. And then it sets
• 3:42
  up a scheduled job to resurrect itself every
• 3:44
  five minutes. Yeah. The source notes one store
• 3:46
  where this resurrection job was added over 1
• 3:48
  ,700 times. An admin would delete it and it would
• 3:51
  just regenerate in less than a second. It's relentless.
• 3:53
  Yeah. But the most alarming detail for anyone
• 3:56
  managing these stores isn't just the profistence.
• 3:59
  It's the fact that patching the software was
• 4:01
  entirely irrelevant to stopping the current infections.
• 4:04
  See, I saw that in the notes and assumed it was
• 4:06
  a typo. How can a fully updated store still be
• 4:10
  compromised? Because the vulnerability allowed
• 4:12
  attackers to plant their persistence mechanisms
• 4:15
  deep in the operating system before the patch
• 4:17
  was even available. Oh, I see. Yeah, stores running
• 4:20
  version 2 .4 .6, patch 15, which is the most
• 4:23
  current patch level they were hit, even a store
• 4:25
  on the newer version 2 .4 .8 running a paid commercial
• 4:29
  protection product got compromised. So patching
• 4:31
  does nothing if they're already in. Exactly.
• 4:33
  If your store was online and exposed between
• 4:36
  September 4th and September 7th, you need a full
• 4:38
  compromise assessment, not just a software update.
• 4:41
  You literally have to assume they are already
• 4:43
  inside. So you have to burn it all down, basically.
• 4:46
  Pretty much. You have to flush user sessions,
• 4:48
  rotate encryption keys, change admin passwords,
• 4:51
  everything, and be warned about the detection
• 4:53
  trap here. What's the trap? If your security
• 4:56
  team just searches for the exact text string
• 4:59
  that the initial security reports published,
• 5:01
  they will fail to find the malware. Because the
• 5:03
  attackers changed the fingerprint. Exactly. The
• 5:05
  infections quickly moved to a different system
• 5:07
  log directory. And that static marker string
• 5:10
  changed within 24 hours. You have to hunt for
• 5:14
  the behavior of the malware, like a hidden process
• 5:16
  constantly restarting, not just scan for a static
• 5:20
  line of code. And this Magento flaw is just,
• 5:23
  it's one piece of a much larger crisis happening
• 5:26
  at the edge of corporate networks right now.
• 5:27
  Oh, yeah. It's everywhere. The source points
• 5:29
  out that the American Catalog of Exploited Vulnerabilities
• 5:32
  added 10 different edge device entries in just
• 5:35
  three days. We're talking about a massive flaw
• 5:38
  in Cisco Secure Firewall Management Center giving
• 5:41
  attackers root access. Yep. Flaws in Citrix Netscaler,
• 5:45
  Fortinet, and this wild microtrick vulnerability
• 5:47
  taking over microtrick routers via SSH without
• 5:50
  any authentication at all. I mean, why are the
• 5:53
  devices meant to protect us suddenly the most
• 5:55
  vulnerable things on the network? It is the cruelest
• 5:58
  irony in modern cybersecurity, really. For years,
• 6:01
  the fundamental rule of defense has been to reduce
• 6:04
  your attack surface. You know, shut down open
• 6:07
  ports, turn off unnecessary services. Right.
• 6:10
  But with a firewall manager or a VPN gateway
• 6:13
  or a load balancer, you cannot reduce their attack
• 6:17
  surface because the attack surface is the product.
• 6:19
  Right. They have to face outward. They are designed
• 6:21
  to stand at the absolute edge of your network
• 6:24
  and face the public internet 24 -7. And because
• 6:28
  these devices are getting exploited before the
• 6:30
  vulnerabilities are even publicly disclosed,
• 6:32
  the fact that you applied a patch today tells
• 6:35
  you absolutely nothing about whether someone
• 6:37
  already walked through that door last week. So
• 6:39
  the strategy has to change. The new mandatory
• 6:41
  rule for IT strategy is patch, then hunt. But
• 6:45
  fixing the software that faces the outside world
• 6:47
  only solves half the problem. If attackers slip
• 6:50
  past that perimeter firewall, what are they walking
• 6:53
  into? Because this week, they are walking into
• 6:56
  a Windows environment that requires nearly a
• 6:58
  thousand separate bandages. It's really bad.
• 7:01
  Let's look at Microsoft's historic September
• 7:03
  2026 patch Tuesday. The volume of updates here
• 7:07
  is staggering. It is a massive payload. I mean,
• 7:09
  the running total for Windows vulnerabilities
• 7:11
  in 2026 is now roughly 2 ,760. Which is insane.
• 7:17
  Let's put that into perspective. The previous
• 7:19
  record year for vulnerabilities was 2020, which
• 7:21
  had 12 ,245. We have more than doubled the worst
• 7:25
  year on record, and we still have a quarter of
• 7:27
  the year left. Mixed into those hundreds of patches
• 7:29
  are two actively exploited zero days. They're
• 7:32
  both Windows privilege escalation flaws carrying
• 7:35
  a CVSS score of 7 .8. Right. I have to push back
• 7:39
  here, or at least ask for some clarity on how
• 7:41
  security teams prioritize these things, because
• 7:43
  we just spent 10 minutes treating a 10 .0 CVSS
• 7:45
  score from Magento like a five alarm fire. Why
• 7:48
  is a 7 .8 being treated with the same extreme
• 7:51
  urgency by these security researchers? Like,
• 7:53
  what does privilege escalation actually mean
• 7:55
  in plain English for a business owner? What's
• 7:57
  fascinating here is that the raw severity score
• 7:59
  can sometimes mask the real world operational
• 8:03
  danger. A 10 .0 is the flaw that gets the attacker
• 8:06
  into the building. But a 7 .8 privilege escalation
• 8:10
  flaw is what they use once they are already inside.
• 8:13
  Think of it like a corporate badge system. A
• 8:16
  phishing email might get an attacker a visitor
• 8:18
  badge, giving them access to the lobby. Right.
• 8:21
  They're in, but they can't do much. Exactly.
• 8:23
  But a privilege escalation flaw is a glitch in
• 8:26
  your badging software that lets that visitor
• 8:28
  walk up to a kiosk and instantly reprint their
• 8:31
  ID to say CEO. Suddenly they have access to the
• 8:34
  server room, the executive suites, everything.
• 8:36
  Okay, that makes total sense. It's the mechanism
• 8:38
  that takes a minor contained breach, like, say,
• 8:41
  a single compromised laptop in HR, and turns
• 8:44
  it into a company -wide ransomware catastrophe.
• 8:46
  Precisely. It gives them system control. And
• 8:49
  while IT teams are triaging that massive list
• 8:51
  of patches, the source highlights two others
• 8:54
  that demand immediate attention. What are those?
• 8:56
  First, an exchange server flaw with an 8 .1 score.
• 9:00
  This triggers remote code execution simply by
• 9:03
  a user receiving a malicious Visio diagram attachment
• 9:06
  in an email. Wow, just receiving it. Yeah. Second,
• 9:09
  a 9 .8 score flaw in the Windows DNS server,
• 9:13
  which researchers are calling the spiritual successor
• 9:15
  to the infamous cigarette vulnerability. DNS
• 9:18
  is basically the phone book of the Internet,
• 9:19
  right? It translates website names into IP addresses.
• 9:22
  Yes. And if you run an Internet -facing Windows
• 9:25
  DNS server, a flaw like this acts like a corrupted
• 9:28
  switchboard operator. An attacker can hijack
• 9:31
  that server and misdirect all your internal and
• 9:34
  external network traffic wherever they want.
• 9:36
  So you think you're going to your bank, but you're
• 9:37
  actually going to their fake site. Exactly. That
• 9:40
  is a drop everything and fix it moment. But there
• 9:43
  is a massive warning from the source material
• 9:45
  regarding this specific patch Tuesday. You absolutely
• 9:48
  must test these updates in a sandbox environment
• 9:50
  before deploying them to your live servers. Yeah,
• 9:53
  I read this part and genuinely felt bad for system
• 9:56
  administrators. The September Windows server
• 9:59
  update actually causes a deadlock between the
• 10:02
  remote desktop protocol and the local session
• 10:05
  manager. I know that sounds super dense, but
• 10:08
  what is the actual mechanism happening there?
• 10:10
  It's essentially a software traffic jam that
• 10:13
  can't be resolved. Think of it like two security
• 10:16
  guards at a single narrow door who both refuse
• 10:19
  to move until the other one steps aside. So they
• 10:22
  just stand there forever. Exactly. Nothing can
• 10:24
  pass. When this deadlock happens, it completely
• 10:27
  breaks remote desktop services across your entire
• 10:30
  environment. Meaning, if an IT director pushes
• 10:32
  this update blindly just to secure the network,
• 10:35
  they accidentally lock their own IT team out
• 10:38
  of the very servers they are trying to manage
• 10:40
  remotely. Exactly. You secure the building by
• 10:42
  supergluing all the locks shut with your team
• 10:45
  on the outside. That's terrifying. You really
• 10:47
  cannot afford to skip the testing phase, even
• 10:50
  when the pressure to patch a thousand vulnerabilities
• 10:52
  is this high. Right. Well, if your software is
• 10:55
  this brittle, you'd hope your hardware provides
• 10:57
  a solid foundation, or at least that your company
• 11:00
  can afford to keep it updated so you can run
• 11:01
  all these new security protocols. But the hardware
• 11:04
  reality check that just hit corporate IT departments
• 11:07
  following Apple's latest event completely rewrites
• 11:10
  that budget math. It was a massive hardware drop,
• 11:12
  sure. But the real story for enterprises is hidden
• 11:16
  in what didn't happen. The event was undeniably
• 11:18
  flashy. I mean, we finally got the foldable iPhone
• 11:21
  Duo opening up to a 7 .8 -inch screen. Pricing
• 11:25
  starting at $1 ,999 and scaling all the way up
• 11:29
  to $3 ,199. Which is a lot for a phone. A whole
• 11:32
  lot. Plus, they rolled out the iPhone 18 Pro
• 11:35
  and Pro Max featuring that new 2 -nanometer A20
• 11:38
  Pro chip alongside a whole suite of new watches
• 11:41
  and AirPods. It's just an incredible amount of
• 11:43
  new tech. If we connect this to the bigger picture,
• 11:45
  though. Those ultra -premium specs are creating
• 11:47
  a logistical nightmare for corporate device fleet
• 11:49
  managers. How so? The critical omission in that
• 11:52
  lineup is that there is no standard iPhone 18
• 11:54
  in this release cycle. Apple has delayed the
• 11:56
  more affordable baseline models to spring 2027.
• 12:00
  So how does a six -month delay on a standard
• 12:02
  phone upend a corporate IT budget? Well, imagine
• 12:05
  you're an IT director for a mid -sized company,
• 12:08
  maybe 1 ,000 employees. You budgeted $800 per
• 12:12
  device for a mid -tier hardware refresh this
• 12:14
  quarter. Your employees' current phones are aging
• 12:18
  out of their security support windows, which,
• 12:21
  as we just discussed with all these flaws, is
• 12:23
  incredibly dangerous right now. You definitely
• 12:26
  want supported devices. Exactly. But because
• 12:28
  the standard model is delayed, your assumptions
• 12:31
  just got destroyed. Your options right now start
• 12:34
  exclusively at the pro -pricing tiers, which
• 12:36
  might be $1 ,200. You are suddenly looking at
• 12:39
  a $400 ,000 deficit in a quarterly budget you
• 12:42
  finalized months ago. Wow. So you either have
• 12:44
  to go to the CFO and beg for half a million dollars
• 12:46
  of unplanned capital expenditure, or you force
• 12:49
  your employees to run vulnerable aging hardware
• 12:52
  for another six to eight months and just pray
• 12:53
  you don't fail a compliance audit. That is the
• 12:56
  exact corner thousands of businesses have just
• 12:58
  been painted into. It's a huge problem. I mean,
• 13:00
  a $400 ,000 hardware deficit is enough to get
• 13:03
  an IT director fired. But when we zoom out to
• 13:06
  the macroeconomic scale of the tech industry
• 13:09
  right now, that kind of money is literally a
• 13:11
  rounding error compared to what is being spent
• 13:14
  to build the infrastructure for artificial intelligence.
• 13:16
  Oh, entirely. It's not even a drop in the bucket.
• 13:18
  I was looking at Oracle's Q1 FY2027 financials
• 13:23
  in the source material, and the stock market's
• 13:25
  reaction made zero sense to me. They posted $19
• 13:28
  .3 billion in total revenue. Their infrastructure
• 13:32
  cloud revenue was up 121%. When the numbers dropped,
• 13:35
  the stock surged almost 8 % during the day. But
• 13:38
  then it completely plummeted and closed 2 % down.
• 13:41
  What spooked the market so badly? It was a single
• 13:44
  line item that revealed the staggering, almost
• 13:47
  unimaginable scale. scale of the AI infrastructure
• 13:49
  boom, and more importantly, the terrifying risk
• 13:51
  holding it up. OK, what was the line item? The
• 13:54
  market initially cheered because Oracle revealed
• 13:56
  a backlog, the remaining performance obligations
• 13:58
  of $664 billion. $664 billion. Yeah. With a B.
• 14:04
  Yes. Heavily driven by $30 billion in new AI
• 14:07
  cloud contracts signed in just one quarter. alongside
• 14:10
  the delivery of 300 ,000 GPUs. But the market
• 14:13
  reversed course violently because they realized
• 14:15
  the extreme counterparty concentration risk.
• 14:18
  Meaning who they're doing business with. Right.
• 14:20
  Roughly half of that massive $664 billion backlog
• 14:24
  is tied to a single entity, OpenAI. Wait, so
• 14:28
  Oracle is basically betting the entire farm on
• 14:31
  one horse. It raises a vital question about whose
• 14:34
  balance sheet is actually underwriting these
• 14:36
  massive physical data centers. Look at the mismatch
• 14:38
  in scale here. OpenAI's estimated annualized
• 14:41
  revenue was around $12 billion. Okay. Yet they
• 14:44
  are making infrastructure commitments in the
• 14:46
  hundreds of billions of dollars. Meanwhile, to
• 14:48
  fund building these physical data centers to
• 14:50
  house all those GPUs for OpenAI, Oracle sold
• 14:53
  $20 billion in stock. They are currently running
• 14:56
  a negative free cash flow of $5 .4 billion, and
• 14:59
  they are guiding a net cash outlay of roughly
• 15:01
  $70 billion for the year. So Oracle is spending
• 15:04
  tens of billions in actual present -day cash
• 15:07
  to pour concrete and buy servers based on a half
• 15:11
  -trillion -dollar IOU from a startup that only
• 15:14
  generates $12 billion a year in revenue. You've
• 15:17
  just identified exactly why the stock dropped.
• 15:19
  If OpenAI's subscriber growth slows down, or
• 15:23
  if the next generation of their models doesn't
• 15:25
  deliver the expected commercial leap, Oracle
• 15:27
  is left holding the bag on tens of billions of
• 15:29
  dollars of highly specialized depreciating Physical
• 15:33
  hardware. That is, I mean, that's insane. It
• 15:35
  is the ultimate lesson in supplier concentration
• 15:37
  risk. And this top -end infrastructure spending
• 15:40
  looks even more precarious when you contrast
• 15:42
  it with the reality of what's happening at the
• 15:44
  bottom end of the AI market. On one hand, you
• 15:46
  have Anthropic reportedly signing compute agreements
• 15:48
  worth up to $517 billion over an 11 -month period.
• 15:53
  consuming 14 .8 gigawatts of power. Which is
• 15:56
  astronomical. Right. To put that in perspective,
• 15:58
  14 gigawatts could power a small country. But
• 16:01
  on the exact same week, a Chinese AI company
• 16:03
  named DeepSeek releases their V4 .1 flash model.
• 16:06
  And the economics of that DeepSeek release fundamentally
• 16:09
  challenged the multi -billion dollar thesis.
• 16:11
  So the source notes it has open weights and an
• 16:13
  MIT license. Let's translate that for a second.
• 16:15
  Open weights means the actual brain of the AI,
• 16:18
  the underlying neural network, can be downloaded
• 16:20
  by anyone. And the MIT license means... a business
• 16:23
  can use it commercially with almost no restrictions.
• 16:25
  Right. It's totally open. And it scores near
• 16:27
  the absolute top frontier level on coding benchmarks.
• 16:31
  But the cost to run it, 60 cents per million
• 16:34
  output tokens, that's 60 cents to generate roughly
• 16:38
  750 ,000 words. The gap between the top end and
• 16:42
  the bottom end of the AI market is, it's no longer
• 16:45
  a gap. It's a chasm. Yeah. The architectural
• 16:48
  efficiency is happening in the open source community,
• 16:50
  utilizing techniques like routing queries to
• 16:53
  smaller, specialized sub -models. Instead of
• 16:55
  activating a massive trillion parameter brain
• 16:57
  for every simple question, they're driving cost
• 17:00
  to zero. Now, there are major caveats, of course.
• 17:03
  Like, well, if you use a model like DeepSeek.
• 17:05
  You have to navigate complex data residency laws.
• 17:08
  If you don't host it yourself, your corporate
• 17:10
  data is bouncing through an overseas API. And
• 17:13
  you actually need the internal engineering talent
• 17:15
  to download those open weights and spin up the
• 17:18
  servers in -house. True, but if a company does
• 17:21
  have that talent, the build versus buy math for
• 17:24
  enterprise AI has totally inverted. I mean, a
• 17:28
  business processing millions of legal documents
• 17:30
  or analyzing massive data sets can suddenly do
• 17:33
  it for absolute pennies using highly capable
• 17:35
  open models rather than paying the astronomical
• 17:38
  frontier infrastructure prices demanded by the
• 17:40
  major labs. Which puts immense existential pressure
• 17:43
  on those heavily leveraged half trillion dollar
• 17:46
  infrastructure bets we just talked about with
• 17:47
  Oracle, OpenAI and Anthropic. The competitive
• 17:51
  landscape and the fundamental value of a generated
• 17:53
  word is shifting beneath their feet while they
• 17:56
  are still busy pouring concrete for their new
• 17:57
  data centers. So what does this all mean? We've
• 18:00
  covered a massive amount of ground today, and
• 18:03
  it really paints a picture of a tech landscape
• 18:04
  experiencing absolute whiplash. In just a few
• 18:08
  days, we are dealing with zero -click Magento
• 18:11
  payloads that completely bypass all software
• 18:14
  patches. We are staring down historically massive
• 18:17
  Microsoft updates that act like a digital traffic
• 18:20
  jam, literally locking IT teams out of their
• 18:22
  own servers. We have Apple's supply chain logistics
• 18:26
  quietly forcing corporate fleets into budget
• 18:28
  -busting hardware. upgrades. And we are watching
• 18:31
  hundreds of billions of dollars being leveraged
• 18:33
  on AI infrastructure that relies on the continued,
• 18:35
  flawless success of a single customer. The throughline
• 18:39
  connecting all of this is that blind trust is
• 18:41
  the single biggest vulnerability a business can
• 18:43
  have today. Whether that is blindly trusting
• 18:45
  that a software patch perfectly protects your
• 18:48
  network perimeter without you actually hunting
• 18:50
  for the intrusion, or blindly trusting an automated
• 18:53
  operating system update won't break your internal
• 18:56
  architecture. or blindly trusting that massive
• 18:59
  AI suppliers have the financial gravity to underwrite
• 19:02
  the future of your company's technology stack.
• 19:05
  The digital perimeter isn't just crumbling. The
• 19:08
  foundation it sits on is actively shifting. You
• 19:10
  have to verify everything. You have to assume
• 19:12
  breach as a baseline operating state and deeply
• 19:15
  understand the mechanics of the tools you rely
• 19:17
  on. Which leaves you with one final thought to
• 19:20
  ponder as we wrap up. We mentioned DeepSeek offering
• 19:22
  near frontier AI capabilities for just 60 cents
• 19:26
  a million tokens, while giants are committing
• 19:28
  hundreds of billions of dollars to physical hardware.
• 19:31
  It raises a massive question about the macroeconomic
• 19:33
  reality of tech right now. If the actual commercial
• 19:36
  value generated by these AI models is being aggressively
• 19:39
  commoditized and replicated by open source alternatives
• 19:43
  for literal pennies, are we staring directly
• 19:46
  at the imminent burst of an AI hardware bubble?
• 19:49
  Because right now, the astronomical cost of building
• 19:52
  the infrastructure seems to be vastly outpacing
• 19:54
  the revenue those models can ever mathematically
• 19:57
  hope to generate. A truly sobering thought to
• 19:59
  end on, and something every tech leader needs
• 20:02
  to factor into their long -term strategy. Absolutely.
• 20:04
  Thanks for joining us for this deep dive into
• 20:06
  the underlying mechanics of our digital world.
• 20:09
  Keep a close eye on your perimeter, test your
• 20:11
  patches before you push them, and we will see
• 20:13
  you next time.