Latest / Tech Talks With Kinsoft / Last Week in Tech – Copilot's Billing Backlash, SoftBank's €75B France Bet, and the Klue OAuth Supply-Chain Hit

Transcript

Last Week in Tech – Copilot's Billing Backlash, SoftBank's €75B France Bet, and the Klue OAuth Supply-Chain Hit · Tech Talks With Kinsoft

• 0:00
  So today, an AI coding assistant just build a
• 0:02
  small startup like $800 just for processing math.
• 0:06
  Yeah, casual $800. Right. While at the same time,
• 0:09
  a 75 billion euro power grid expansion is literally
• 0:13
  being mapped out in France just to keep that
• 0:16
  kind of math running. It's massive. And on top
• 0:19
  of all that, Wikipedia volunteers are actually
• 0:21
  striking because human workers are being laid
• 0:24
  off to pay for it all. So, welcome to the Deep
• 0:27
  Dive. Glad to be here. We are so glad to have
• 0:30
  you with us today. We've got a stack of sources
• 0:32
  to go through, anchored by a really revealing
• 0:34
  briefing from Kinsoft Tech Talks. It's dated
• 0:37
  June 8th, 2026. Right. The Kinsoft briefing.
• 0:40
  Yeah. And our mission today for this deep dive
• 0:43
  is to look at the severe structural reality check
• 0:47
  that is hitting the tech industry right now because
• 0:49
  we are definitely moving past that honeymoon
• 0:51
  phase of the seamless AI -powered upgrade. Oh,
• 0:54
  the honeymoon is definitely over. Totally over.
• 0:56
  We're entering this era of steep, multifaceted
• 1:00
  consequences. So we're going to track the shock.
• 1:03
  of metered ai billing models yeah the actual
• 1:06
  physical limits of global infrastructure and
• 1:08
  and this is a big one how our rush to integrate
• 1:11
  all these powerful tools has created just catastrophic
• 1:14
  blind spots in corporate security it's a really
• 1:17
  critical set of sources to synthesize i mean
• 1:20
  because it forces us to look at the hidden causality
• 1:23
  of the tools we use every single day right because
• 1:26
  we tend to treat software as you know infinite
• 1:29
  and weightless like it's just magic in the cloud
• 1:31
  exactly magic but But what this briefing exposes
• 1:34
  is that we are slamming into hard physical and
• 1:36
  economic limits. The entire tech ecosystem is
• 1:39
  currently scrambling to re -architect itself
• 1:42
  before those limits completely break the current
• 1:45
  business models. Yeah. So let's start with that
• 1:48
  economic limit, which really hits right at the
• 1:49
  developer's wallet. Where it hurts. Exactly.
• 1:52
  The Kinsoft briefing details what happened on
• 1:54
  June 1st. So GitHub changed the billing structure
• 1:57
  for its AI coding assistant, Kotilot, for a really
• 2:00
  long time. This was just a flat monthly subscription.
• 2:04
  Twenty nine dollars. Nice and easy. Super predictable.
• 2:07
  Right. You put it on the corporate card and you
• 2:09
  just forget about it. Yeah. But then they flipped
• 2:11
  it to a token based pay as you use billing model.
• 2:14
  Yeah. And I mean, the reaction was explosive.
• 2:17
  Oh, people were furious. Yeah, because developers
• 2:20
  were opening their dashboards expecting that
• 2:23
  standard twenty nine dollar charge. And instead
• 2:25
  they're seeing. bills for hundreds of dollars
• 2:28
  a month. Yeah, which fundamentally shatters the
• 2:31
  financial predictability that, honestly, the
• 2:34
  entire software as a service model has been built
• 2:36
  on for the last 20 years. Right. Okay, let's
• 2:39
  unpack this. Because, you know, if you're listening,
• 2:41
  think about it like this. It's like going to
• 2:43
  your favorite all -you -can -eat buffet. Okay,
• 2:46
  I like buffets. Right. You pay your 20 bucks
• 2:48
  at the door, you eat as much as you want. But
• 2:50
  suddenly management decides to change the rules
• 2:53
  and now they're charging you per individual bite
• 2:55
  you take. That is a fast way to ruin a dinner.
• 2:58
  Exactly. How are everyday businesses supposed
• 3:00
  to budget for these tools when AI powered suddenly
• 3:04
  means it's metered? Well, when you shift from
• 3:07
  a flat sauce license to metered token billing,
• 3:11
  you are exposing the end user to the raw fluctuating
• 3:16
  cost of compute. But I want to push back on that
• 3:18
  justification a little bit because I get it.
• 3:20
  AI inference is expensive. Sure. Right. But we've
• 3:23
  had cloud computing and massive database lookups
• 3:26
  for. decades, right? And we didn't suddenly start
• 3:29
  billing users per keystroke. So why does generating
• 3:32
  a line of code in an IDE suddenly require like
• 3:35
  a taxi meter? That's a great question. It's because
• 3:38
  the mechanics of inference are entirely different
• 3:40
  from traditional compute. And that's the core
• 3:42
  issue GitHub is actually trying to offset here.
• 3:44
  Okay. How so? So when you execute a traditional
• 3:47
  database query, the server locates the data,
• 3:49
  retrieves it, and the transaction is done. Super
• 3:51
  efficient. Highly efficient. Yeah. But when an
• 3:54
  AI assistant generates code, it isn't just...
• 3:56
  looking up an answer in a filing cabinet, it
• 3:58
  is continuously processing tokens. And tokens
• 4:01
  are what, like fragments of words? Exactly. Fragments
• 4:04
  of words or bits of code. It's processing those
• 4:07
  through billions of parameters. And to do that,
• 4:10
  the model has to be loaded into the incredibly
• 4:12
  fast, highly constrained VRAM of specialized
• 4:16
  GPUs. Right. So it's monopolizing hardware in
• 4:20
  a way a standard web request never does. Exactly.
• 4:22
  For the entire duration that the AI is streaming
• 4:25
  that code back to your screen, it is locking
• 4:27
  up prime real estate on a GPU cluster somewhere.
• 4:31
  Wow. It's performing continuous matrix multiplication.
• 4:34
  And that requires an amazing... So when GitHub
• 4:38
  was charging just $29 a month. They were heavily
• 4:42
  subsidizing the variable cost of that math. I
• 4:44
  mean, it was a loss leader to get developers
• 4:46
  addicted to the workflow. OK, that makes sense.
• 4:49
  But now that the tool is deeply embedded in enterprise
• 4:51
  pipelines, you know, they're passing the actual
• 4:53
  literal cost of hardware utilization directly
• 4:56
  onto the user. Which creates a massive operational
• 4:58
  nightmare for anyone running a business. Oh,
• 5:00
  absolutely. I mean, if I'm a CTO at a startup
• 5:03
  and AI powered. Suddenly means I have this uncapped
• 5:07
  metered liability running in the background of
• 5:09
  every single developer's machine. I can't forecast
• 5:12
  my burn rate. You're flying blind. Yeah. Like
• 5:15
  one month my team is doing light bug fixes and
• 5:18
  the bill is 50 bucks. But the next month we have
• 5:22
  a massive crunch to ship a feature. The AI is
• 5:25
  generating massive blocks of context and suddenly
• 5:28
  my infrastructure bill is $1 ,000 higher. And
• 5:31
  that unpredictability is the exact vulnerability
• 5:33
  that the big tech giants are desperately trying
• 5:36
  to patch right now. Because they know it's going
• 5:38
  to scare people off. Exactly. They know metered
• 5:40
  billing will stall enterprise adoption. In fact,
• 5:43
  the Kinsoft source notes that at Microsoft's
• 5:45
  Build Conference, they spent an enormous amount
• 5:47
  of time showcasing homegrown, smaller coding
• 5:50
  models. Oh, interesting. Yeah, they are specifically
• 5:53
  designing these to be cheaper to run than the
• 5:55
  big name generalized state of the art models.
• 5:58
  But wait, are those cheaper models? actually
• 6:00
  any good? Well, that's the debate. Because, look,
• 6:03
  if I'm a developer relying on Copilot to catch
• 6:06
  security flaws or write really complex back -end
• 6:09
  logic, I don't want some lobotomized, lightweight
• 6:13
  AI writing my code just because Microsoft wants
• 6:16
  to save margins on their GPU clusters. Right.
• 6:18
  I mean, isn't the industry just... Offering a
• 6:21
  substandard product to fix a billing problem
• 6:23
  they created themselves. That is the multibillion
• 6:26
  dollar tension right now. I mean, they are aggressively
• 6:28
  pushing small language models or SLS. Right.
• 6:31
  Under the premise that you don't need a trillion
• 6:35
  parameter model that knows the history of the
• 6:37
  Roman Empire just to write a simple Python script.
• 6:40
  Fair point. So by training highly specialized
• 6:42
  narrow models, they can reduce the VRAM requirements
• 6:46
  and execute the inference much faster and much,
• 6:49
  much cheaper. But it's risky, right? You are
• 6:51
  absolutely right to be skeptical because the
• 6:53
  boundary of where a small model fails and, you
• 6:56
  know, hallucinates is much narrower. The industry
• 6:58
  is really trying to thread the needle here. They
• 7:00
  have to drop the compute cost so they don't bankrupt
• 7:03
  their users with token fees, but they can't degrade
• 7:06
  the product so much that users just abandon it
• 7:08
  completely. Right. And the reason they are so
• 7:11
  desperate to drop that compute cost brings us
• 7:14
  to the second major theme of the briefing. Because
• 7:17
  the scarcity that's driving these token prices
• 7:19
  up, it isn't just server availability. No, not
• 7:22
  at all. It's the physical limits of the planet's
• 7:25
  infrastructure. Yeah. We have to look at the
• 7:27
  staggering scale of capital being deployed just
• 7:29
  to keep this math running. So the Kinsoft briefing
• 7:33
  highlights this announcement from SoftBank. This
• 7:36
  is a huge one. It's wild. They have committed
• 7:39
  up to 75 billion euros to build five gigawatts
• 7:42
  of AI data center capacity in France. It is almost
• 7:45
  impossible to overstate how massive five gigawatts
• 7:48
  of power consumption actually is. Put that in
• 7:51
  perspective for us. Well, to ground that in reality,
• 7:53
  a typical large scale nuclear reactor generates
• 7:55
  about one gigawatt of power. Wait, really? One
• 7:58
  gigawatt per reactor. Yeah. Yes. So SoftBank
• 8:01
  isn't just building a standard server farm. They
• 8:04
  are proposing an energy footprint equivalent
• 8:06
  to five nuclear power plants. That is just wow.
• 8:10
  And it's dedicated entirely to the matrix multiplication
• 8:13
  required to train and run these models. Here's
• 8:16
  where it gets really interesting. That explains
• 8:18
  why they are targeting France, specifically.
• 8:21
  Exactly. Because if you need 5 gigawatts of continuous,
• 8:24
  unbroken baseline power, you can't rely entirely
• 8:28
  on intermittent renewables. You need the robust
• 8:31
  nuclear grid that France has built. You need
• 8:33
  that stability. Yeah, but think about the sheer
• 8:35
  volume of concrete, steel, cooling systems, and
• 8:39
  grid -level transformers required to build a
• 8:41
  5 gigawatt footprint. It's mind bending. Yeah.
• 8:45
  I mean, we always talk about the cloud as if
• 8:46
  it's weightless. Like it's just floating up there.
• 8:48
  Right. But this is a 75 billion euro industrial
• 8:51
  megaproject. If we connect this to the bigger
• 8:54
  picture, it's the most aggressive mobilization
• 8:57
  of private capital into physical infrastructure
• 8:59
  we've seen in decades. Yeah. What we are witnessing
• 9:02
  is the collision between exponential software
• 9:05
  demands and linear physical reality. That's a
• 9:08
  great way to put it. Because you can update software
• 9:10
  overnight, right? But you cannot build a gigawatt
• 9:14
  substation overnight. Definitely not. The supply
• 9:16
  chains for industrial cooling, for high voltage
• 9:19
  transformers and for the specialized cabling
• 9:22
  required for these data centers, they're backlogged
• 9:25
  by years. So that's why the tokens are so expensive.
• 9:28
  Exactly. The token prices we discussed earlier
• 9:31
  are high precisely because this physical infrastructure
• 9:34
  is fundamentally bottlenecked. And while 75 billion
• 9:37
  euros is flowing into concrete and copper wire,
• 9:40
  the Kinsaw briefing contrasts that with a deeply
• 9:43
  sobering human. This part is really tough. It
• 9:47
  is. Over at Wikipedia, their volunteer editors
• 9:50
  actually went on strike. Yeah. And the core issue,
• 9:53
  they were protesting layoffs among the paid staff,
• 9:55
  which they blamed directly on AI driven cost
• 9:58
  cutting. It is a perfect distillation of the
• 10:01
  broader capital shift happening across the entire
• 10:03
  economy right now. Well, on one side of the ledger,
• 10:06
  you have unprecedented fortunes being poured
• 10:09
  into energy grids and GPU clusters. And on the
• 10:12
  exact opposite side, you have the human. element
• 10:14
  being systematically squeezed out just to free
• 10:18
  up the capital to buy that compute. I mean, the
• 10:20
  irony with Wikipedia is incredibly thick here.
• 10:23
  Yeah. You have a platform that represents the
• 10:25
  absolute pinnacle of collaborative human curated
• 10:28
  knowledge. The best of the Internet, really.
• 10:30
  Yeah. And the volunteers, people who dedicate
• 10:33
  their time for free are having to strike to protect
• 10:36
  the paid administrative roles. It's backwards.
• 10:38
  They're being displaced by the exact same generative
• 10:41
  AI models that were literally trained. on the
• 10:44
  free labor of those very same Wikipedia contributors.
• 10:46
  Yep. It feels less like an upgrade and more like
• 10:49
  an extraction, to be honest. That's a highly
• 10:51
  accurate way to frame the mechanics of what's
• 10:53
  happening. When an organization automates a human
• 10:56
  role with an AI tool, the savings rarely just
• 10:59
  sit in a bank account. Right. They spend it.
• 11:01
  They spend it immediately. That capital is almost
• 11:04
  instantly reallocated into purchasing more software
• 11:07
  licenses, more compute power, and more token
• 11:10
  usage. Wow. The Wikipedia strike is a clear indicator.
• 11:14
  that the infrastructure arms race is not happening
• 11:16
  in a vacuum. It is being directly subsidized
• 11:19
  by the restructuring of human labor. I mean,
• 11:23
  we are literally transferring wealth from payrolls
• 11:27
  to power grids. Which is a very heavy macroeconomic
• 11:30
  reality. But I want to bring this back down to
• 11:33
  the operational level for you listening right
• 11:34
  now, because there is an immediate... technical
• 11:37
  fallout to this massive AI adoption. And it's
• 11:40
  messy. Super messy. As organizations adopt these
• 11:43
  expensive tools and desperately try to squeeze
• 11:46
  productivity out of them to justify that, we
• 11:51
  wire the AI into our CRMs, our email servers,
• 11:55
  our cloud drives. We just want all the tools
• 11:57
  talking to each other. Seamless workflow. Exactly.
• 11:59
  But the third pillar of the Kinsoft briefing
• 12:01
  makes it very clear that this obsession with
• 12:03
  seamless integration has created a catastrophic
• 12:06
  security landscape. So we really need to talk
• 12:08
  about OAuth supply chain attacks. This is undoubtedly
• 12:11
  the most critical and actionable threat vector
• 12:14
  mentioned in the sources. The attack surface
• 12:16
  of the modern enterprise has completely shifted.
• 12:19
  OK, so anyone managing a SOSDAQ knows OAuth.
• 12:23
  It's that protocol that allows you to click.
• 12:25
  integrate with Salesforce or connect to Google
• 12:28
  Workspace. We all use it. It's incredibly convenient
• 12:31
  because it allows third -party apps to talk to
• 12:33
  your central databases without you having to
• 12:35
  hand over your actual master password. Right.
• 12:38
  But the briefing details a major incident with
• 12:41
  a sales intelligence company called Clue. Right.
• 12:43
  And Clue is a practically legitimate platform
• 12:45
  that sales teams use to aggregate data. But the
• 12:48
  attackers didn't breach Clue because they cared
• 12:50
  about Clue's internal memos or anything. No,
• 12:52
  they didn't care about Clue's data. They breached
• 12:55
  Clue to get access to the Oeth integrations that
• 12:58
  Clue's customers had authorized. And by hijacking
• 13:01
  those integrations, the attackers were able to
• 13:03
  siphon highly sensitive Salesforce data directly
• 13:06
  out of the systems of Clue's customers. And this
• 13:09
  requires us to look really closely at how Oeth
• 13:12
  scopes are actually implemented in the wild.
• 13:14
  Okay, bring that down for us. So when a user
• 13:16
  clicks allow access on one of those pop -ups,
• 13:20
  they are generating an authorization token. Right.
• 13:23
  That token dictates exactly what the third -party
• 13:26
  app is allowed to do. That's its scope. The fatal
• 13:29
  flaw here is that the industry has a really bad
• 13:31
  habit of blindly over -provisioning these scopes.
• 13:35
  Oh, like giving them too much permission? Way
• 13:37
  too much. A basic analytics plugin might request
• 13:40
  full read and write access to an entire CRM database,
• 13:44
  and a rushed employee will just click approve
• 13:47
  to get the pop -up out of the way and get back
• 13:48
  to work. So what does this all mean? To visualize
• 13:51
  the mechanism here, let's look at how hotel key
• 13:53
  cards work. Okay. When you check into a hotel,
• 13:56
  they don't give you the physical master key to
• 13:58
  the building, right? They program a temporary
• 14:01
  key card that is supposed to only open your specific
• 14:03
  room for three days. Right. Very limited scope.
• 14:05
  But what we're doing with OAuth is the equivalent
• 14:08
  of an employee going to the front desk, programming
• 14:11
  a key card that inexplicably grants permanent
• 14:13
  access to the penthouse, the security room, and
• 14:16
  the financial vault. And then handing that key
• 14:18
  card over to a third -party dog walker or vendor
• 14:21
  just because the vendor asked nicely. That's
• 14:23
  exactly it. If that vendor's pocket gets picked,
• 14:26
  the burglar doesn't have to break your windows
• 14:28
  or bypass your alarms. They just swipe the perfectly
• 14:31
  valid key card you created for them. And because
• 14:33
  it is a valid cryptographically signed token,
• 14:36
  it completely bypasses your multi -factor authentication.
• 14:40
  Wait, really? It bypasses MFA? Yes, it bypasses
• 14:44
  MFA. It bypasses your biometric logins. The system
• 14:47
  assumes that because the token is legitimate,
• 14:50
  the traffic is legitimate. That is terrifying.
• 14:53
  This is the very definition of a supply chain
• 14:55
  attack. The attackers target a mid -tier vendor
• 14:58
  with weaker security specifically to harvest
• 15:01
  the high -level tokens belonging to their much
• 15:03
  larger, much more secure enterprise clients.
• 15:06
  Which brings me to the detail in the briefing
• 15:08
  that is just staggering. The victim list. Yes.
• 15:11
  The victims of this clue breach. The companies
• 15:14
  that had their core Salesforce data exfiltrated
• 15:16
  included several dedicated security firms. Yeah,
• 15:19
  that's the part that hurts. How is that even
• 15:22
  possible? These are companies selling zero -trust
• 15:25
  architectures and advanced threat detection.
• 15:27
  Don't they audit their own integrations? What's
• 15:30
  fascinating here is it perfectly illustrates
• 15:32
  the concept of shadow IT and the total collapse
• 15:35
  of the traditional security perimeter. Shadow
• 15:37
  IT meaning stuff happening under the radar. Exactly.
• 15:40
  In a modern organization, it is rarely the IT
• 15:44
  department actually provisioning these OAuth
• 15:47
  tokens. Right. It's a marketing manager connecting
• 15:49
  a new SEO tool or a sales director integrating
• 15:53
  a productivity app. Just trying to get their
• 15:55
  job done. Right. So the security firms were compromised
• 15:58
  because the technical enforcement of their perimeter
• 16:01
  was bypassed by human convenience. They trusted
• 16:04
  a vendor, the vendor got compromised, and the
• 16:07
  blast radius traveled directly down the supply
• 16:09
  chain right into the security firm's database.
• 16:12
  It renders the entire concept of a firewall almost
• 16:16
  irrelevant. Pretty much. You were only as secure
• 16:18
  as the weakest marketing plug -in your intern
• 16:20
  connected three years ago and forgot about. That's
• 16:23
  the reality. Yeah. And the Kinsoft briefing emphasizes
• 16:26
  that this is not an isolated incident. They explicitly
• 16:29
  tie it to the repeating pattern of the Shiny
• 16:32
  Hunters threat group, who utilized identical
• 16:35
  methods to execute massive Salesforce thefts.
• 16:39
  It's a known playbook now. And to show just how
• 16:41
  relentless this broader perimeter collapse is,
• 16:44
  the source notes two other major breaches from
• 16:47
  that exact same week. Which ones? Nintendo getting
• 16:50
  hit by ransomware and losing employee data and
• 16:53
  Oxford University's careers platform being breached
• 16:55
  to expose student details. Oh, wow. I mean, whether
• 16:58
  it's global gaming giants or ancient universities,
• 17:01
  the data is just bleeding out from every conceivable
• 17:04
  vector. It's the inevitable result of hyper connectivity.
• 17:07
  I mean, we. have prioritized seamless API communication
• 17:10
  above literally everything else. Right. But every
• 17:13
  single one of those API connections, every OAuth
• 17:16
  token is a potential doorway. Just waiting to
• 17:18
  be opened. Exactly. The attackers have realized
• 17:21
  that trying to batter down the front door of
• 17:24
  a highly secure enterprise is a waste of time.
• 17:26
  It is infinitely more efficient to quietly slip
• 17:30
  through the side doors that were just left propped
• 17:32
  open by third party integrations. Which brings
• 17:34
  us to the synthesis of what this June. June 2026
• 17:38
  briefing is really telling us today. Yeah, bringing
• 17:41
  it all together. We started by looking at the
• 17:43
  economic reality of the tools we use, right?
• 17:45
  The shift to unpredictable metered token billing
• 17:49
  that forces users to absorb the raw cost of compute.
• 17:52
  And the sticker shock that comes with it. Then
• 17:54
  we traced why that compute is so expensive. Because
• 17:57
  it requires the deployment of massive gigawatt
• 18:00
  scale physical infrastructure that is actively
• 18:03
  displacing human labor in the pursuit of capital
• 18:05
  efficiency. The Wikipedia strike being the prime
• 18:07
  example. Exactly. And finally, we looked at how
• 18:10
  our rush to integrate these powerful, expensive
• 18:12
  systems has basically weaponized our own connectivity
• 18:16
  against us through Outh supply chain attacks.
• 18:18
  The overarching narrative across all these sources,
• 18:21
  if you look at them together, is friction. Friction,
• 18:24
  yeah. The era of frictionless tech adoption is
• 18:26
  over. Every computational choice now carries
• 18:30
  a distinct, measurable weight, whether that weight
• 18:33
  is measured in euros, in human jobs, or in security
• 18:36
  vulnerabilities. Which means we have to move
• 18:38
  from observation to action. The practical takeaway
• 18:41
  from listening to this right now is immediate.
• 18:44
  You need to audit your digital blast radius.
• 18:47
  Do it today. Yes, literally today. Log into your
• 18:50
  core platforms, whether that's Google Workspace,
• 18:53
  Microsoft 365, GitHub, Salesforce, whatever you
• 18:56
  use, navigate to the security settings, and pull
• 18:59
  the list of connected third -party applications.
• 19:01
  You might be surprised by what you find. You
• 19:03
  are almost guaranteed to find integrations that
• 19:06
  were authorized years ago by employees who may
• 19:08
  not even work there anymore. If an app doesn't
• 19:10
  have a strict operational necessity today, Revoke
• 19:14
  its token immediately. Revoke it. You have to
• 19:16
  aggressively deprecate the lingering access keys
• 19:18
  you've left scattered across the supply chain.
• 19:20
  That kind of aggressive digital hygiene is really
• 19:23
  the only defense when the perimeter itself is
• 19:26
  decentralized? But, you know, looking at the
• 19:28
  entirety of what we've unpacked today, it raises
• 19:30
  a really profound structural question. Oh, what's
• 19:34
  that? Well, we are rapidly moving into an environment
• 19:37
  where every interaction with an interface might
• 19:40
  incur a metered token charge, right? Where the
• 19:42
  underlying infrastructure requires city -sized
• 19:45
  power grids, and where simply connecting two
• 19:48
  apps exposes your core data to global supply
• 19:51
  chain attacks. It's exhausting just thinking
• 19:53
  about it. Right. So if being plugged in... means
• 19:56
  absorbing constant financial and security liabilities,
• 19:59
  will the ultimate luxury in future technology
• 20:02
  be the ability to completely disconnect and execute
• 20:05
  your work locally entirely offline? Wow, that
• 20:08
  is a brilliant angle. If connectivity becomes
• 20:10
  a liability and cloud compute becomes a metered
• 20:13
  utility bill, perhaps the most premium feature
• 20:16
  a piece of software can offer is isolation. Isolation
• 20:19
  as a luxury. Yeah, because as the Kintop briefing
• 20:21
  makes abundantly clear, the plumbing connecting
• 20:24
  our digital world is expensive, it is leaking,
• 20:27
  and the bill has finally arrived. It definitely
• 20:29
  has. Thanks for taking the deep dive with us
• 20:31
  today. Take a hard look at your integrations,
• 20:34
  and we'll see you next time.