Latest / Connecting the Dots / AI-Powered Cybersecurity, Alphabet's AI Ambitions, and Trillion-Dollar Tech IPOs

Transcript

AI-Powered Cybersecurity, Alphabet's AI Ambitions, and Trillion-Dollar Tech IPOs · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarfl Solutions
• 0:06
  Group brings you Connecting the Docs, where we
• 0:09
  break down what's happening in tech and why it
• 0:11
  matters to you. Now, let's dive in with your
• 0:14
  hosts. Imagine a computer program that can...
• 0:18
  You know, look at the digital vault holding your
• 0:20
  life savings, find a microscopic crack that human
• 0:24
  engineers have missed for a decade, and then
• 0:26
  just automatically write the exact code needed
• 0:28
  to break that vault wide open on its very first
• 0:31
  try. Right. And it succeeds 83 % of the time.
• 0:34
  Exactly. And we are not talking about some theoretical
• 0:38
  sci -fi scenario here. We are talking about,
• 0:40
  well, Tuesday morning, welcome to Connecting
• 0:42
  the Dots. I'm Alex. And I'm Morgan. Broadcasting
• 0:44
  live from Snarfle Solutions Group right here
• 0:46
  in Sacramento, California. Every day we take
• 0:48
  the massive wall of tech news coming at you and
• 0:51
  we break down exactly what it means for your
• 0:53
  work, your family, and your future. Yeah, and
• 0:57
  today is Tuesday, June 2nd, 2026, and the entire
• 1:01
  landscape of global security is fundamentally
• 1:03
  shifting beneath our feet today. It really is.
• 1:06
  But before we jump into the heavy stuff, let's
• 1:08
  get you situated with what's happening outside
• 1:10
  and in the markets. Right. So right now in Sacramento,
• 1:13
  it is looking like a gorgeous sunny 93 degrees
• 1:16
  Fahrenheit today with winds blowing from 2 to
• 1:20
  12 miles per hour. Perfect weather. Yeah, it
• 1:22
  is. But if you're traveling or listening from
• 1:25
  elsewhere in the country, just be aware there
• 1:28
  are currently 248 active weather alerts nationally.
• 1:31
  Oh wow, that's a lot. Yeah, and 40 of those are
• 1:34
  overland, so definitely check weather .gov for
• 1:36
  your local details. Good tip. Moving over to
• 1:38
  the financial side, on Monday, June 1st, the
• 1:41
  Dow Jones closed slightly down at 51 ,078 .88,
• 1:45
  so that's a 0 .16 % drop, while the S &P 500
• 1:50
  closed up slightly at 7 ,599 .96. And for the
• 1:54
  crypto watchers out there, Bitcoin took a bit
• 1:56
  of a hit today. It closed down 3 .39 % at 68
• 2:00
  ,905 .72. Ouch. And we also have to acknowledge
• 2:04
  the sports world really quick. The SF Giants
• 2:06
  just got... absolutely crushed by the Milwaukee
• 2:10
  Brewers yesterday. Oh man, don't remind me. 16
• 2:12
  to 2. Yeah, brutal. Chad Patrick took the win
• 2:14
  on the mound, and he frankly pitched a master
• 2:16
  class. He really did. Well, at least the A's
• 2:18
  and the Kings had the day off, so you get a breather
• 2:20
  on that front. Yeah, thank goodness. So let's
• 2:22
  look closer at today's Morningstar Tech Report.
• 2:25
  We're tracking three major developments today
• 2:27
  that really center around this massive theme
• 2:29
  of scale and security in the AI era. Yeah, we
• 2:32
  are basically watching artificial intelligence
• 2:35
  completely dismantle our current understanding
• 2:37
  of digital security, which is forcing the tech
• 2:40
  industry to rebuild the locks on the entire internet.
• 2:44
  And doing so is requiring a level of physical
• 2:47
  infrastructure spending that, I mean, it frankly
• 2:50
  rivals the industrial revolution. It's wild.
• 2:52
  The speed at which this is evolving is hitting
• 2:54
  critical velocity and the implications. They
• 2:57
  touch everything from your local hospitals database
• 2:59
  all the way to the global banking system. Yeah,
• 3:02
  it's huge. But before we impact the first development,
• 3:04
  we have a quick message. Today's episode is brought
• 3:07
  to you by Strong Coffee, the premium instant
• 3:09
  coffee that tastes like a traditional brew but
• 3:11
  is packed with 15 grams of protein, MCTs, and
• 3:14
  adaptogens like ashwagandha. Get clean, sustained
• 3:17
  energy without the jitters. Use promo code SNARFL
• 3:20
  for 20 % off at strongcoffeecompany .com. All
• 3:23
  right, so the first major move happening today
• 3:25
  comes from Anthropic, and it's regarding the
• 3:27
  rapid expansion of something called Project Glasswing.
• 3:30
  Okay, so what is Glasswing doing now? Well, Anthropic
• 3:33
  is taking their highly advanced cybersecurity
• 3:35
  focused AI model, specifically, Claude Mitho's
• 3:39
  preview, and they are pushing it out to about
• 3:42
  150 new organizations. Wow, 150. Yeah, across
• 3:46
  more than 15 countries. OK, but to really grasp
• 3:48
  the weight of this, we have to look at who is
• 3:50
  actually on that list, because Anthropic isn't
• 3:53
  just handing this over to, like, random software
• 3:55
  developers. Right, no. They are deploying this
• 3:58
  AI into critical infrastructure. We are talking
• 4:01
  about the organizations that run power grids,
• 4:04
  municipal water supplies, national health care
• 4:06
  networks, global communications cables. Exactly.
• 4:09
  Physical hardware manufacturing. You have entities
• 4:12
  like Samsung, SK Hynix, NATO, and even the Five
• 4:16
  Eyes Intelligence Alliance plugging this model
• 4:19
  directly into their systems. And the motivation
• 4:21
  behind this massive expansion, it really comes
• 4:24
  down to what Anthropic learned during their pilot
• 4:27
  phase. Back in early April, they gave access
• 4:31
  to a much smaller cohort of roughly 50 initial
• 4:35
  partners. And what did those 50 partners find?
• 4:38
  Well, they turned the AI loose on their own code
• 4:40
  basis just to see what it could find. And the
• 4:42
  results were a massive wake up call to the industry.
• 4:45
  In just a few weeks, the AI uncovered more than
• 4:48
  10 ,000 high or critical severity security flaws.
• 4:52
  10 ,000? Yeah, that human security teams had
• 4:55
  completely missed. 10 ,000 critical flaws. And,
• 4:58
  you know, a critical flaw usually means an attacker
• 5:00
  could take complete remote control of the system
• 5:02
  without even needing a password. Exactly. It's
• 5:05
  terrifying. So to put this in perspective for
• 5:07
  you, it's like giving the neighborhood watch
• 5:08
  a highly advanced AI -powered metal detector
• 5:12
  or an architectural X -ray that shows every single
• 5:15
  faulty deadbolt, hidden tunnel, and broken window
• 5:18
  in the entire city. Right, which is an incredible
• 5:20
  tool for the good guys to have. It is. But if
• 5:23
  this tool is so good at finding vulnerabilities,
• 5:26
  isn't it dangerous to spread it to 150 different
• 5:29
  organizations? I mean, handing out copies of
• 5:32
  this X -ray machine to that many entities drastically
• 5:36
  increases the surface area for a leak. Oh, absolutely.
• 5:39
  Like, eventually a bad actor is going to get
• 5:41
  their hands on this tech, right? Well, that tension
• 5:43
  right there is the defining problem of modern
• 5:46
  cybersecurity. Anthropix leadership, they know
• 5:49
  the risk of proliferation. Yeah. But they are
• 5:52
  operating against a ticking clock. Because of
• 5:54
  other AI models. Exactly. The intelligence they're
• 5:57
  gathering suggests a very specific near -term
• 6:00
  horizon. Like within the next six to twelve months
• 6:03
  the open market is going to be flooded with other
• 6:05
  AI models that has similar Capabilities to mythos
• 6:08
  and those competitors aren't going to have anthropic
• 6:11
  safety guardrails No, most of them won't you're
• 6:13
  gonna have open source models or state -sponsored
• 6:16
  models that get released without any ethical
• 6:18
  restrictions whatsoever So when cheap fast an
• 6:21
  incredible powerful offensive AI hits the open
• 6:25
  web in less than a year The attackers will automatically
• 6:28
  find those 10 ,000 broken locks anyway. So the
• 6:31
  defenders basically need a permanent advantage
• 6:33
  right now. Exactly. The people maintaining your
• 6:36
  bank's routing software or your local traffic
• 6:38
  grid need an asymmetric advantage today to patch
• 6:42
  those holes before the offensive AI wave even
• 6:44
  hits. But finding 10 ,000 flaws is great, I guess,
• 6:48
  but it creates a massive backlog of unpatched
• 6:51
  software, which begs the question... Who exactly
• 6:54
  is allowed to know about these critical vulnerabilities
• 6:57
  once the AI finds them? Let's explore this, because
• 7:00
  someone has to actually write the code to patch
• 7:02
  them. Right, and when you uncover 10 ,000 critical
• 7:06
  flaws across the globe's digital infrastructure,
• 7:08
  it completely breaks the old corporate models
• 7:10
  of secrecy. Because a bank can't just quietly
• 7:13
  fix a flaw if the broken software is made by
• 7:16
  a third -party vendor that, you know, a hundred
• 7:18
  other banks use. Precisely. Which brings us to
• 7:22
  the massive policy shift Anthropic announced
• 7:24
  today. Previously, those 40 to 50 early partners
• 7:28
  and project last -wing groups like Apple, Google,
• 7:30
  Microsoft, the Pentagon, they were bound by strict
• 7:33
  non -disclosure agreements. So they had to keep
• 7:36
  the findings secret. Yeah. If the AI found a
• 7:39
  flaw, they could only report it back to Anthropic.
• 7:42
  It was a completely closed loop. But today, Anthropic
• 7:46
  is tearing up that playbook. Wow. They are now
• 7:49
  allowing these partners to share vulnerability
• 7:51
  findings outward to other security teams, industry
• 7:54
  regulators, open source maintainers, and the
• 7:57
  press under responsible disclosure norms. And
• 8:00
  the underlying reason they're being forced to
• 8:02
  open this up is that 83 % statistic we mentioned
• 8:04
  at the very top of the deep dive. This AI model
• 8:07
  is terrifyingly good. It really is. Anthrobic
• 8:10
  ran internal tests to see just how capable Mythos
• 8:12
  really is. They asked it to find zero -day vulnerabilities,
• 8:16
  which are flaws in major operating systems that
• 8:18
  literally no human on Earth knows about yet.
• 8:20
  And the model found thousands of them, but it
• 8:23
  didn't just point them out. It wrote, working
• 8:26
  weaponized exploits to attack those flaws. And
• 8:29
  it succeeded on the first attempt, 83 % of the
• 8:32
  time. To understand why that is so terrifying,
• 8:36
  you have to look at how hacking traditionally
• 8:37
  works. I mean, a human security researcher might
• 8:41
  spend six months reverse engineering a piece
• 8:43
  of software. just looking for a tiny memory leak.
• 8:46
  Then they spend another three months trying to
• 8:48
  chain that memory leak into a larger failure
• 8:50
  that grants them access. But Mythos is looking
• 8:54
  at the entire architecture of the code, understanding
• 8:57
  the logic flaws instantly and generating the
• 9:00
  attack sequence in seconds. I mean, if an AI
• 9:03
  can compromise global financial infrastructure
• 9:05
  that quickly, keeping it a secret between a tech
• 9:07
  company and a few beta testers is just a recipe
• 9:10
  for systemic collapse. Exactly. Why do you think
• 9:12
  listeners should care about the specific policy
• 9:14
  shift. Because if the Bank of England or the
• 9:16
  European Central Bank doesn't know about a flaw
• 9:19
  in the global financial infrastructure, your
• 9:22
  family's money could literally be at risk. That's
• 9:24
  why regulators push so hard for this transparency.
• 9:27
  And furthermore, the Pentagon is actively using
• 9:29
  mythos right now to patch the US government's
• 9:33
  software. They're trying to harden the grid.
• 9:35
  Right. Regulators realize that if a critical
• 9:37
  flaw exists in the payment routing system used
• 9:40
  by half the planet, You cannot legally or ethically
• 9:44
  keep that information locked inside a private
• 9:46
  beta test. But wait, if they share these findings
• 9:49
  with the public and the press, aren't they basically
• 9:52
  publishing a treasure map for hackers? Like,
• 9:55
  if I'm a cyber criminal, I don't even need to
• 9:56
  build my own AI. I just read the morning news
• 9:59
  to find out exactly which database at which bank
• 10:01
  was left wide open. Well, that is exactly why
• 10:04
  they are implementing a framework called Responsible
• 10:06
  Disclosure. It's a highly structured process.
• 10:09
  When a partner uses mythos and finds a flaw,
• 10:12
  they do not just tweet the code out. Under responsible
• 10:15
  disclosure, they privately alert the company
• 10:18
  that built the software first. And they provide
• 10:20
  a specific embargo window, usually about 90 days,
• 10:24
  for that company's engineers to write, test,
• 10:28
  and push out a software update to all their users.
• 10:30
  So the lock gets fixed in the dark before the
• 10:33
  robbers arrive? Yes. And only after the 90 days
• 10:36
  when the patch is available do they publish the
• 10:38
  finding to the broader industry. And even then,
• 10:41
  they withhold the weaponizable details. They
• 10:44
  explain the nature of the vulnerability so other
• 10:46
  defenders can learn from it, but they strip out
• 10:48
  the actual exploit code that Mythos generated.
• 10:51
  Got it. But, you know, discovering these flaws,
• 10:54
  writing patches, and running simulations with
• 10:56
  these massive AI models, it requires an unfathomable
• 10:59
  amount of computing power. Unfathomable. And
• 11:02
  paying for all that computing power brings us
• 11:03
  to our final story today. Let's look closer at
• 11:06
  the sheer cost of this AI arms race, because
• 11:09
  the math happening in the background of these
• 11:11
  AI models demands physical server farms the size
• 11:14
  of small cities. Yeah, we are seeing capital
• 11:16
  deployment on a historic level right now. Alphabet,
• 11:20
  the parent company of Google, just announced
• 11:23
  they're raising a staggering $80 billion. $80
• 11:27
  billion. Yeah, through a package of equity offerings.
• 11:30
  And that money is earmarked entirely for funding
• 11:33
  their AI infrastructure and custom chips. Let's
• 11:37
  break down the mechanics of an $80 billion capital
• 11:40
  raise, because the way they are getting this
• 11:42
  cash tells you everything you need to know about
• 11:44
  the current market. Right. This isn't just a
• 11:46
  bank loan. Alphabet is using three distinct financial
• 11:49
  vehicles. First, a $40 billion at the market
• 11:53
  offering. Right. This means Alphabet is slowly
• 11:56
  creating new shares and selling them directly
• 11:59
  into the daily trading volume of the stock market.
• 12:02
  Every time a retail investor buys Google stock,
• 12:04
  the cash goes straight to Alphabet's war chest.
• 12:06
  And second, they have $30 billion in underwritten
• 12:10
  stock offerings. That's when massive investment
• 12:13
  banks buy huge blocks of shares to distribute
• 12:15
  to their institutional clients. And finally,
• 12:19
  there's a massive $10 billion private placement
• 12:22
  directly from Warren Buffett's Berkshire Hathaway,
• 12:25
  which is run by Greg Abel. This brings Berkshire's
• 12:28
  total stake in Alphabet to 16 .6%. billion dollars.
• 12:33
  It's just staggering. The end result of this
• 12:36
  80 billion dollar cash injection is that Alphabet's
• 12:39
  annual capital spending forecast has been revised
• 12:42
  to somewhere between 180 billion and 190 billion
• 12:46
  dollars in a single year. Yes. To really grasp
• 12:50
  why a tech company needs to spend almost $200
• 12:52
  billion on infrastructure in one year, we have
• 12:55
  to look at the business realities here. Look
• 12:57
  at their Q1 numbers for 2026. Google services,
• 13:01
  which is their traditional search engine and
• 13:02
  ad business, generated $89 .6 billion in revenue
• 13:06
  with a profit of $40 .6 billion. It's essentially
• 13:09
  a money printer. Right. The cost to serve you
• 13:11
  an ad is practically zero. Exactly. But contrast
• 13:14
  that with Google Cloud, which is driving all
• 13:15
  this AI growth. In that same quarter, Google
• 13:18
  Cloud pulled in 20 billion in revenue, but only
• 13:20
  6 .6 billion in profit. So the margins are way
• 13:23
  lower. Fastly lower because the physical cost
• 13:26
  of running the service is so incredibly high.
• 13:29
  One of our source analysts used this brilliant
• 13:30
  analogy comparing this to Berkshire Hathaway's
• 13:33
  history with C's candies and the BNSF Railway.
• 13:37
  Oh, I like this one. Yeah. Think of Google's
• 13:39
  ad business like a candy store. It prints money
• 13:42
  with zero effort. The profit margins are huge.
• 13:46
  But what do you do with tens of billions of dollars
• 13:49
  in candy profits? You take that cash and you
• 13:52
  buy the incredibly expensive railroad tracks
• 13:55
  of the future right because the BNSS railway
• 13:57
  is the exact opposite of a candy store It is
• 14:00
  heavily industrialized. It requires constant
• 14:02
  massive injections of capital just to lay steel
• 14:05
  tracks and buy locomotives Yes, but the barrier
• 14:07
  to entry is so unimaginably high that once you
• 14:10
  own the tracks You own the future of commerce
• 14:13
  in that region. Alphabet is basically taking
• 14:15
  the effortless cash from their digital candy
• 14:17
  store and using it to lay the railroad tracks
• 14:20
  of the AI economy, which are these AI data centers
• 14:22
  and compute. Exactly. This contextualizes why
• 14:26
  Berkshire Hathaway is investing so heavily. AI
• 14:29
  compute is becoming the ultimate commodity. Cash
• 14:32
  is king in this race. I mean, look at what happened
• 14:34
  recently when SpaceX had excess computing power.
• 14:36
  Oh, right. They leased it out to Entropic at
• 14:39
  an incredibly high premium. The market set an
• 14:42
  astronomical price for compute, and Anthropic
• 14:44
  just paid it because supply is so constrained.
• 14:47
  Alphabet is signaling that consumer and enterprise
• 14:50
  demand for AI is vastly exceeding supply, and
• 14:53
  they are willing to do whatever it takes to build
• 14:55
  out the physical infrastructure. Absolutely.
• 14:57
  Before we wrap up, a quick thanks to Old Glory,
• 15:00
  an iconic music and sports fan merch store, perfect
• 15:03
  for the techie in your life. Use promo code SNARFL
• 15:06
  for 15 % off at oldglory .com. So, to briefly
• 15:10
  summarize how all these threads connect back
• 15:13
  to your daily life, we are watching the real
• 15:15
  -time rebuilding of global cybersecurity through
• 15:17
  AI. Yeah. We're seeing the creation of completely
• 15:20
  new rules for how tech secrets are shared, and
• 15:23
  we're witnessing the deployment of historic amounts
• 15:25
  of money to physically power this new era of
• 15:28
  technology. It's an incredible time to be watching
• 15:30
  this space, but it leaves me with a final provocative
• 15:34
  thought for you to ponder today. If models like
• 15:37
  Mythos can autonomously find and patch thousands
• 15:41
  of flaws, and hyperscalers like Google are spending
• 15:44
  hundreds of billions to run them, will there
• 15:47
  eventually be a future where human software engineers
• 15:50
  are no longer responsible for writing or defending
• 15:52
  code at all? That's a crazy thought. What happens
• 15:55
  when AI is both the ultimate hacker and the ultimate
• 15:58
  security guard? That really is the question we
• 16:00
  have to answer over the next decade. We're here
• 16:02
  to help at Snarfl, so reach out on our website
• 16:04
  if you have any more questions. That's a wrap
• 16:09
  on today's top stories. Remember, tech doesn't
• 16:12
  have to be complicated, and we're here to help
• 16:14
  connect the dots. If you've got questions or
• 16:17
  want to learn more, visit us at snarful .com.
• 16:21
  Thanks for listening, and we'll catch you next
• 16:23
  time.