Latest / Tech Talks With Kinsoft / Last Week in Tech – Nvidia's Record Quarter, OpenAI's $110B Raise, and Washington Blacklists Anthropic

Transcript

Last Week in Tech – Nvidia's Record Quarter, OpenAI's $110B Raise, and Washington Blacklists Anthropic · Tech Talks With Kinsoft

• 0:00
  Imagine raising $110 billion and then actually
• 0:04
  refusing to take a large chunk of it in cash.
• 0:07
  Yeah, it sounds completely backward, doesn't
• 0:09
  it? It really does. But, I mean, we are looking
• 0:12
  at a technology landscape right now where...
• 0:14
  The U .S. dollar is, you know, it's starting
• 0:16
  to take a backseat to a brand new global currency.
• 0:19
  And that currency is compute. Exactly. And if
• 0:22
  you're listening to this, you already know that
• 0:24
  keeping up with artificial intelligence feels,
• 0:26
  well, it feels like trying to drink from a fire
• 0:30
  hose while you're riding a bullet train. That
• 0:31
  is a very accurate description. Right. So you
• 0:34
  don't need another summary of just basic headlines.
• 0:37
  You need to understand the underlying teconic
• 0:40
  plates that are shifting beneath those headlines.
• 0:42
  So today we are jumping. jumping straight into
• 0:44
  the mechanics of what is actually happening.
• 0:46
  And we really have the perfect roadmap for this
• 0:48
  today. Yeah, we do. We are dissecting an incredibly
• 0:51
  dense, highly revealing tech digest from Tech
• 0:54
  Talks with Kinsoft. And this is dated March 2nd,
• 0:57
  2026. Our mission for this deep dive is to untangle
• 1:01
  this wild paradox playing out right in front
• 1:04
  of us. We're going to explore how unprecedented
• 1:06
  just astronomical financial investments in AI
• 1:09
  are directly triggering these complex geopolitical
• 1:12
  standoffs. Which is wild. on its own. Right.
• 1:14
  But how simultaneously those very same advanced
• 1:17
  technologies are actively arming everyday cyber
• 1:21
  criminals, you know, to scale up these terrifyingly
• 1:24
  mundane threats. OK, let's unpack this, because
• 1:27
  we have to start with the sheer scale of the
• 1:29
  money. The money is just staggering. It really
• 1:31
  is. The Kinsoft Digest, it leads with a financial
• 1:34
  reality that frankly. I mean, it breaks traditional
• 1:37
  economic models. They report that NVIDIA posted
• 1:40
  a record quarter of around $68 billion. Yeah,
• 1:44
  $68 billion. Just to be clear, that is a single
• 1:46
  quarter, up over 70%, and driven almost entirely
• 1:49
  by data -centered chips for AI. Which is massive.
• 1:52
  But right on the heels of that, we get the real
• 1:54
  shocker. OpenAI closed a massive $110 billion
• 1:57
  funding round, and that was led by Amazon, NVIDIA,
• 2:00
  and SoftBank. But... The detail that, you know,
• 2:04
  really demands attention is that much of Nvidia's
• 2:09
  share in that open AI funding didn't come as
• 2:13
  a wire transfer of cash. It came as compute.
• 2:16
  Which is a fundamental restructuring of how tech
• 2:19
  equity works. What's fascinating here is that
• 2:23
  the processing power itself, like the raw capability
• 2:26
  to train these massive models, has... completely
• 2:30
  bypass traditional currency. It's crazy to think
• 2:33
  about. It is. It's now the primary medium of
• 2:35
  exchange because, I mean, think about it. If
• 2:37
  you are open AI, a billion dollars in cash is
• 2:39
  nice. Sure. But you're just going to take that
• 2:41
  cash, go to a vendor and buy processing power
• 2:44
  anyway. Right. You're going to buy the chip.
• 2:45
  Exactly. So NVIDIA is essentially cutting out
• 2:48
  the middleman here. They're taking equity in
• 2:50
  exchange for dedicated server racks, processing
• 2:52
  time and, you know, silicon access. Okay, wait,
• 2:55
  I have to push back here for a second because
• 2:57
  this does sound familiar to me. Oh, really? Yeah,
• 2:59
  you're saying compute is this revolutionary new
• 3:01
  currency. But, I mean, tech companies have always
• 3:04
  traded in hardware and server space. That's true.
• 3:06
  So it's kind of like, okay, think of it like
• 3:08
  a lumber company, right? A lumber company funding
• 3:11
  a massive housing development, not by writing
• 3:13
  a check to the developer, but by simply providing
• 3:16
  all the wood they could ever possibly need. Right,
• 3:19
  I see where you're going with this. So how is
• 3:21
  NVIDIA doing this today fundamentally? Any different
• 3:24
  from, say, IBM in the 1990s providing mainframes
• 3:28
  to enterprise startups at a discount in exchange
• 3:31
  for equity? It's a really fair question. But
• 3:34
  the difference lies in the nature of the asset
• 3:37
  itself and the macroeconomics at play. Okay.
• 3:40
  How so? When IBM provided a mainframe, they were
• 3:42
  providing a tool to run a business. You know,
• 3:44
  a very expensive tool, but still just a tool.
• 3:46
  Right. In the AI ecosystem, compute isn't just
• 3:49
  a tool. It is the raw, consumable commodity that's
• 3:53
  actually required to synthesize the intelligence
• 3:55
  itself. Ah, I see. A much better way to understand
• 3:58
  this is to look at the early days of the petrodollar.
• 4:00
  Wait, the petrodollar? As in oil. Exactly. Think
• 4:04
  about how the global economy shifted when oil
• 4:06
  became the undeniable bedrock resource required
• 4:09
  for industrialization. Oh, wow. Right. Control
• 4:12
  over the raw energy supply dictated international
• 4:15
  trade agreements. It dictated equity and national
• 4:18
  security. Compute is essentially the new oil.
• 4:22
  That's a huge comparison. But it fits perfectly.
• 4:25
  NVIDIA isn't just handing over hardware. They're
• 4:28
  allocating access to a scarce, globally demanded
• 4:31
  resource that operates as the foundational energy
• 4:35
  of the entire AI economy. So they're locking
• 4:38
  them in. Yes. When a company invests compute,
• 4:40
  they are locking the recipient into a highly
• 4:43
  specific architectural ecosystem. On a balance
• 4:46
  sheet, it allows NVIDIA to secure equity in the
• 4:49
  very companies that will drive their future chip
• 4:51
  demand. And they do it without depleting their
• 4:53
  cash reserves. Oh, man. So it's a closed loop
• 4:56
  economic engine. Exactly. So it essentially guarantees
• 4:58
  that NVIDIA's technology is the inescapable foundation
• 5:01
  of OpenAI's future developments. You're tying
• 5:04
  the most advanced AI company in the world directly
• 5:06
  to your specific hardware architecture. And that
• 5:09
  level of capital investment leads us to a fascinating
• 5:12
  transition. Yeah, because when you're throwing
• 5:14
  around numbers like $110 billion and you're dealing
• 5:17
  in entire alternative currencies based on processing
• 5:20
  power, you aren't just a nimble tech startup
• 5:22
  operating in a vacuum anymore, are you? Not at
• 5:25
  all. And this is where the causality becomes
• 5:27
  totally undeniable. When a technology requires
• 5:31
  the gross domestic product of a small country
• 5:33
  just to build its physical infrastructure, it
• 5:37
  crosses an invisible threshold. It really does.
• 5:39
  It stops being just a commercial software product.
• 5:42
  It becomes foundational global infrastructure.
• 5:45
  These companies begin to operate with the weight
• 5:47
  and influence of nation states. Yeah, that makes
• 5:49
  sense. And when you possess nation state levels
• 5:51
  of power and you control a resource that rivals
• 5:54
  oil, you inevitably crash headfirst into government
• 5:58
  intervention. I mean, you cannot amass $110 billion
• 6:00
  in strategic infrastructure without the federal
• 6:03
  government taking a very close, very critical
• 6:06
  look at how you operate. Which brings us directly
• 6:09
  to the second major revelation in this March
• 6:12
  2026 Kinsoft Digest. And I want to frame this
• 6:17
  carefully because it is a stark operational shift.
• 6:19
  It really is. According to the Digest, the Trump
• 6:22
  administration took a very aggressive regulatory
• 6:25
  stance regarding a major AI company called Anthropic.
• 6:29
  The administration officially blacklisted Anthropic,
• 6:32
  ordering federal agencies to immediately stop
• 6:35
  using their technology. And the terminology they
• 6:38
  use there is critical. Yeah. They didn't just
• 6:40
  cancel a contract. They applied a highly consequential
• 6:43
  label to the company. The government officially
• 6:46
  labeled Anthropic a supply chain risk. Right.
• 6:49
  And the source material explains the catalyst
• 6:51
  for this action. This blacklisting occurred after
• 6:54
  Anthropic refused to drop its internal corporate
• 6:57
  limits on two specific use cases. Which were?
• 7:00
  Autonomous weapons and mass surveillance applications
• 7:02
  of its AI. Right. Now, as we explore this, it
• 7:06
  is absolutely crucial for you, the listener,
• 7:08
  to know that we are looking at this strictly
• 7:10
  from a market reality perspective. We are not
• 7:13
  here to debate the politics of the administration's
• 7:16
  mandate. And we are not here to weigh in on the
• 7:18
  moral philosophy of Anthropic's internal ethics
• 7:21
  committee. Our focus is entirely on the shockwaves
• 7:25
  this event sends through the business world.
• 7:27
  Because those shockwaves are massive. Precisely.
• 7:30
  Because regardless of where anyone stands on
• 7:32
  the political or ethical spectrum of that specific
• 7:35
  standoff, the operational reality for the market
• 7:38
  shifted practically overnight. Yeah, no kidding.
• 7:41
  The digest highlights a profound new reality.
• 7:45
  How an AI vendor navigates government pressure
• 7:48
  and internal ethics is no longer an abstract
• 7:51
  philosophical debate. Right. It is an immediate
• 7:53
  practical business liability for any customer
• 7:56
  relying on their technology. Here's where it
• 7:58
  gets really interesting. I mean, this completely
• 8:00
  rewrites the textbook on what a supply chain
• 8:03
  risk actually means. It does. If you are a chief
• 8:05
  operating officer and you hear supply chain risk,
• 8:08
  you traditionally think of physical, logistical
• 8:10
  bottlenecks. Right, like a cargo ship. Yeah.
• 8:13
  You think of a cargo ship stuck sideways in the
• 8:16
  Suez Canal or a global shortage of semiconductor
• 8:19
  manufacturing or, you know, a factory shutting
• 8:22
  down because of a hurricane. Exactly. You understand
• 8:24
  how to model and mitigate physical delays. But
• 8:27
  this is entirely different. If a software company's
• 8:30
  internal philosophical ethical guidelines can
• 8:34
  trigger a sudden federal blacklist, doesn't that
• 8:36
  mean your entire software stack is vulnerable
• 8:38
  to a new kind of sudden death? If we connect
• 8:41
  this to the bigger picture, it validates exactly
• 8:44
  what you were asking. The definition of a supply
• 8:46
  chain has evolved entirely. It really has. In
• 8:49
  the digital age, your supply chain is the software
• 8:51
  you rent, it's the cloud compute you rely on,
• 8:54
  and the AI models you integrate into your daily
• 8:56
  operations. Yeah. What we are seeing in this
• 8:59
  snapshot is the realization that ideological
• 9:01
  and ethical friction is now a tangible, immediate
• 9:04
  liability. Let me put you in the shoes of a business
• 9:07
  leader, then. Okay. If I'm a logistics CEO, say,
• 9:09
  running a massive supply chain network moving
• 9:12
  freight across North America, How am I supposed
• 9:15
  to underwrite the risk of my software vendor's
• 9:17
  internal ethics committee? That is the big question.
• 9:20
  Right. Because I chose a vendor to optimize my
• 9:24
  shipping routes. I spent millions integrating
• 9:26
  their API into my dispatch systems. I have absolutely
• 9:30
  nothing to do with autonomous weapons or federal
• 9:33
  surveillance. Obviously not. But suddenly, because
• 9:36
  of a geopolitical standoff over those issues,
• 9:39
  my vendor is blacklisted. Is it even possible
• 9:43
  to insure against that kind of risk? That is
• 9:47
  the multi -million dollar question that's keeping
• 9:49
  chief risk officers awake at night right now.
• 9:51
  I bet. Traditionally, you know, you insure against
• 9:53
  a data breach or a server outage. Insuring against
• 9:56
  regulatory sudden death caused by a vendor's
• 9:59
  moral stance is completely uncharted territory.
• 10:01
  Unbelievable. If your logistics business handles
• 10:04
  any federal subcontracts or interacts with federal
• 10:06
  data systems, which almost all major logistics
• 10:09
  firms do, you might suddenly be forced to rip
• 10:12
  out your entire AI infrastructure overnight just
• 10:15
  to remain compliant. Why? You were completely
• 10:17
  caught in the crossfire. So you have to audit
• 10:19
  not just the technical capability and security
• 10:21
  of your vendors, but their political positioning,
• 10:24
  their lobbying efforts, and their ethical red
• 10:26
  lines. Yep. You are essentially... betting on
• 10:30
  the geopolitical diplomacy skills of your software
• 10:33
  provider just to ensure your own shipping software
• 10:35
  keeps running. Exactly. You are no longer just
• 10:38
  buying a product. You are adopting their geopolitical
• 10:40
  footprint. That is wild. But, you know, as massive
• 10:44
  as these top tier nation state level standoffs
• 10:47
  are, the Kinsoft Digest doesn't let us stay in
• 10:50
  the stratosphere. Right. It brings us back down.
• 10:52
  It forces us to look at how this immense technological
• 10:55
  capability bleeds down into the gritty, ground
• 10:59
  -level reality of day -to -day operations. Yeah.
• 11:02
  Because while governments and tech titans are
• 11:04
  fighting over billion -dollar compute allocations
• 11:06
  and ethical frameworks, the exact same AI technology
• 11:10
  is actively being weaponized in the trenches.
• 11:13
  Which brings us to the cybersecurity desk portion
• 11:15
  of the digest. And frankly, this is where the
• 11:18
  operational reality gets incredibly chilling.
• 11:20
  It really does. Let's lay out the facts reported
• 11:22
  for that single week in March. First. A Russian
• 11:26
  -speaking threat actor was caught using generative
• 11:29
  AI to compromise hundreds of Fortinet firewalls
• 11:32
  globally. Just to quickly define that for context,
• 11:35
  Fortinet firewalls are essentially the digital
• 11:37
  bouncers standing between the company's internal
• 11:40
  corporate network and the open public internet.
• 11:44
  Good analogy. Thanks. And the attackers got in
• 11:46
  through weak management interfaces, which is
• 11:49
  basically like leaving the bouncer's own internal
• 11:51
  rulebook and master keys sitting on a table out
• 11:54
  in the open. It is a devastating oversight, but
• 11:57
  unfortunately, a really common one. And the report
• 11:59
  continues from there. Right. So on top of the
• 12:01
  Fortinet breaches, researchers discovered AI
• 12:04
  -generated code actively being used to help orchestrate
• 12:07
  a cyber attack on Mexican government systems.
• 12:10
  And to cap it all off, a known hacker crew called
• 12:13
  Shiny Hunters struck again. They hit a major
• 12:17
  car marketplace called CarJurus, exposing over
• 12:20
  12 million users. And they simultaneously stole
• 12:23
  employee data from a massive entity like Wynn
• 12:26
  Resorts. It is just a relentless drumbeat of
• 12:29
  breaches across wildly different sectors. It
• 12:31
  is relentless, but the crucial insight from the
• 12:33
  Kinsoft report isn't merely that hacks are happening.
• 12:36
  As you know, hacks always happen. That isn't
• 12:39
  news. The critical insight is the mechanism of
• 12:42
  how AI is fundamentally changing the nature of
• 12:44
  these attacks. The digest makes a very specific
• 12:47
  point. Attackers are now using AI to scale up
• 12:50
  the boring, effective stuff. So what does this
• 12:54
  all mean? Because I think when a lot of us hear
• 12:56
  about AI and cybersecurity, our minds immediately
• 12:59
  go to Hollywood tropes. Oh, absolutely. We picture
• 13:01
  a hyper -intelligent, autonomous program actively
• 13:05
  fighting our security software in a digital war
• 13:07
  room. We imagine hackers deploying some incredibly
• 13:11
  complex, never before seen zero day virus. Right.
• 13:15
  You know, a vulnerability no one even knew existed
• 13:17
  that just effortlessly outsmarts human defenders.
• 13:21
  But that is entirely disconnected from what this
• 13:23
  report is actually saying. Yes. The reality is
• 13:26
  much more mundane, which ironically makes it
• 13:28
  far more dangerous to the average organization.
• 13:30
  How so? The threat isn't a science fiction super
• 13:32
  virus. It is extreme automation. I want to make
• 13:35
  sure we really nail the mechanics of this. In
• 13:38
  the past, if we talked about hackers scanning
• 13:40
  the Internet for vulnerabilities, we usually
• 13:42
  compared it to a metal detector. Yeah, that's
• 13:44
  the standard analogy. Right. Hackers used automated
• 13:46
  scripts to just walk the digital beach, sweeping
• 13:49
  back and forth endlessly, waiting for a beep
• 13:51
  that told them someone left a port open. Right.
• 13:53
  But you and I both know that traditional non
• 13:57
  -AI automated scripts, things like NMAP, have
• 14:00
  been doing that for decades. That isn't new.
• 14:03
  So how does generative AI actually change this
• 14:06
  dynamic? This raises an important question, and
• 14:09
  it is the exact distinction we need to make.
• 14:11
  Traditional tools like NMAP are just metal detectors.
• 14:15
  They find a potential target and they alert the
• 14:18
  human hacker. The human then has to sit down.
• 14:21
  analyze the specific system, figure out its unique
• 14:23
  configuration, and manually write or adapt a
• 14:27
  piece of exploit code to break in. Right. It
• 14:29
  takes work. Exactly. It takes time, expertise,
• 14:31
  and a lot of manual effort. Okay. So if the old
• 14:34
  way is just the metal detector, what is the generative
• 14:37
  AI doing differently? Generative AI upgrades
• 14:40
  that process exponentially. It's like having
• 14:42
  a metal detector that doesn't just beep when
• 14:44
  it finds a buried treasure chest, but it instantly
• 14:46
  analyzes the lock. And right there on the spot,
• 14:48
  3D prints the exact customized key needed to
• 14:53
  open it. Oh, wow. Yeah. The AI isn't just scanning.
• 14:56
  It is dynamically writing the exploit payload
• 14:59
  on the fly to bypass specific localized defenses
• 15:02
  once it finds a vulnerability. That is, I mean,
• 15:05
  so it completely removes the bottleneck of human
• 15:08
  labor. Entirely. In the past, a hacker might
• 15:10
  find a thousand vulnerable firewalls, but only
• 15:13
  have the time and manpower to actively exploit
• 15:16
  10 of them before the systems got patched. Right.
• 15:18
  Now the AI can orchestrate the scanning of thousands
• 15:21
  of endpoints, instantly write custom scripts
• 15:24
  for each unique environment, and automatically
• 15:26
  execute the exploit to grab the data. All simultaneously
• 15:29
  while the human attacker is literally asleep.
• 15:32
  Precisely. It is the complete automation of payload
• 15:34
  creation and execution. That's terrifying. And
• 15:37
  that ties directly back to the specific warning
• 15:39
  highlighted in the Kinsoft Digest. Never expose
• 15:42
  a firewall or appliances admin interface to the
• 15:45
  open internet. And operate under the assumption
• 15:47
  that attackers are using AI to find the ones
• 15:50
  that are. If a management interface is left even
• 15:53
  slightly exposed anywhere on the public internet,
• 15:55
  an AI -powered sweep won't just find it. It will
• 15:59
  exploit it before a human IT team even logs in
• 16:02
  for their morning coffee. Which perfectly explains
• 16:04
  the sheer volume and variety of the attacks we
• 16:06
  just listed. It explains how a crew like Shiny
• 16:09
  Hunters can hit a niche car marketplace to expose
• 16:12
  12 million users and then immediately pivot to
• 16:15
  stealing internal employee data from a massive
• 16:18
  hospitality conglomerate like Wynn Resorts. The
• 16:21
  scope or industry of the target just doesn't
• 16:23
  matter anymore. When the AI is doing the heavy
• 16:25
  lifting of dynamically exporting vulnerabilities,
• 16:28
  it ceases to be a targeted operation. It is just
• 16:31
  a frictionless numbers game. It lowers the barrier
• 16:34
  to entry for threat actors to near zero. to be
• 16:37
  elite coders to attack a complex system they
• 16:40
  just need an ai model capable of interpreting
• 16:43
  the exposed interface and generating the right
• 16:46
  commands wow And this is exactly why the Kinsoft
• 16:49
  Digest ends with that very direct, almost aggressive
• 16:52
  question aimed at IT administrators. Is our firewall
• 16:56
  management locked down? Yeah. Because if you
• 16:58
  cannot answer that question instantly with absolute
• 17:01
  certainty, you're already behind. That makes
• 17:03
  total sense. The attackers have outsourced their
• 17:05
  reconnaissance and their weaponization to AI,
• 17:08
  meaning your window of time to patch a known
• 17:10
  vulnerability before it is discovered has effectively
• 17:13
  shrunk to zero. It really puts this entire deep
• 17:16
  dive into a stark perspective. When you take
• 17:18
  a step back and look holistically at this single
• 17:20
  isolated week in March 2026, the contrast in
• 17:24
  how this technology permeates our world is just
• 17:27
  staggering. It really is. We started at the absolute
• 17:29
  pinnacle of global finance, looking at a technology
• 17:32
  fueled by $110 billion funding rounds, where
• 17:35
  the physical processing power, the compute, is
• 17:37
  so incredibly valuable that it has bypassed cash
• 17:40
  to become a literal currency among tech titans.
• 17:42
  And then we followed that immense value and scale
• 17:45
  to its logical conclusion, which is fact. Right.
• 17:48
  We saw how building infrastructure on that scale
• 17:51
  transforms the technology into a highly politicized
• 17:55
  geopolitical asset. The internal corporate ethics
• 17:58
  of a company regarding weapons and surveillance
• 18:01
  can trigger sudden federal blacklists, completely
• 18:04
  redefining what a supply chain risk looks like
• 18:07
  for any everyday business relying on those systems.
• 18:10
  Right. The technology is so advanced and so powerful,
• 18:12
  it forces presidential administrations to step
• 18:14
  in and draw red lines. Yes. And then at the exact
• 18:17
  same moment, that exact same technological leap
• 18:20
  is being leveraged by ground level cyber criminals.
• 18:23
  It's a crazy juxtaposition. It is. And not to
• 18:26
  launch sophisticated sci -fi cyber wars against
• 18:29
  intelligence agencies, but to quietly and ruthlessly
• 18:32
  automate the mundane process of finding weak
• 18:35
  management interfaces on firewalls just to steal
• 18:38
  employee data and user credentials from car websites.
• 18:42
  It is a profound realization for you as the listener
• 18:44
  trying to navigate this landscape. Whether you
• 18:47
  are actively managing an investment portfolio,
• 18:50
  whether you're a chief operating officer trying
• 18:53
  to secure a complex digital supply chain against
• 18:56
  sudden regulatory blacklists, or whether you
• 19:00
  are just an everyday consumer hoping your personal
• 19:02
  data on a retail marketplace is secure, these
• 19:05
  three pillars are now permanently intertwined.
• 19:08
  Capital, politics, and mundane cybersecurity.
• 19:12
  all completely upended by the exact same tool
• 19:14
  at the exact same time. You cannot separate the
• 19:18
  $110 billion valuation from the geopolitical
• 19:21
  blacklisting. And you cannot separate the blacklisting
• 19:24
  from the AI -powered firewall breaches. There
• 19:27
  are all symptoms of the same massive uncontrolled
• 19:30
  acceleration. Which leaves us with a critical
• 19:32
  question to consider as we wrap up our analysis
• 19:34
  of this Kinsoft report. Okay, let's hear it.
• 19:37
  We've seen the sheer scale of the investment
• 19:39
  where compute dictates equity and billions are
• 19:41
  moved in a single round. We've seen the severe
• 19:43
  geopolitical and operational vulnerabilities
• 19:45
  that scale creates for businesses. Right. And
• 19:48
  we've seen how effortlessly this technology arms
• 19:50
  everyday threat actors to dynamically exploit
• 19:53
  mundane weaknesses at a global scale. It is a
• 19:56
  lot to process, but it is exactly why we do these
• 19:59
  deep dives. So the question you have to ask yourself
• 20:01
  as we look toward the future of this industry
• 20:03
  is this. If AI is simultaneously our most heavily
• 20:07
  funded foundational technology, a heavily politicized
• 20:10
  and highly regulated government asset, and a
• 20:13
  totally frictionless tool for everyday cybercriminals.
• 20:16
  Well, how long until the financial liability
• 20:18
  of securing these massive systems inevitably
• 20:20
  outplaces the staggering profits of deploying
• 20:22
  them?