Latest / Connecting the Dots / DeepSeek’s New AI Model, EA’s $55B Buyout, and Oracle’s AI Cloud Deal

Transcript

DeepSeek’s New AI Model, EA’s $55B Buyout, and Oracle’s AI Cloud Deal · 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 Dots, 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. Good morning, everyone, and welcome back
• 0:17
  to Connecting the Dots. It is Monday, September
• 0:19
  29th, 2025. I'm Alex. And I'm Morgan. Good morning.
• 0:22
  Ready to connect some dots today? Always. Let's
• 0:26
  start with a quick look outside here in Sacramento.
• 0:29
  Looks like maybe some chance showers today, high
• 0:31
  around 74 degrees Fahrenheit. Yeah, fairly mild
• 0:34
  for us. Nationally, though, sounds like the East
• 0:37
  Coast is still dealing with some rough seas,
• 0:38
  you know, rip currents from Imelda and Humberto.
• 0:42
  So heads up if you're out there. Definitely.
• 0:44
  OK, quick market check before we dive in. How
• 0:46
  did things close out last week? Right. So looking
• 0:48
  at Friday's close, the Dow Jones finished up
• 0:51
  0 .32 % on the day. For the week, it was just
• 0:55
  slightly positive, up 0 .09%. OK, pretty flat
• 0:58
  week for the Dow then. Yeah, basically. The S
• 1:00
  &P 500 gained 0 .43 % on Friday, but it actually
• 1:04
  dipped a tiny bit for the week down 0 .16%. And
• 1:08
  Bitcoin? Bitcoin's seeing small losses right
• 1:10
  now, down 0 .17 % today, and it's off about 0
• 1:14
  .66 % over the past week. So maybe market's just
• 1:18
  sort of stabilizing or maybe reacting to some
• 1:20
  of the tech news we're about to cover. Could
• 1:21
  be. All right, before we get to that, let's talk
• 1:23
  about our sponsor for today's show. Yes! We're
• 1:26
  happy to partner with Pins and Aces. They're
• 1:27
  a fantastic premium golf apparel and accessory
• 1:30
  store. Seriously, great stuff. If you're looking
• 1:33
  to upgrade your golf wardrobe, maybe get some
• 1:36
  cool head covers or polos, Pins and Aces is definitely
• 1:38
  the place. And here's the best part. Use our
• 1:41
  promo code SNARFUL, that's S -N -A -R -F -U -L,
• 1:46
  SNARFUL at checkout, and you'll get 21 % off
• 1:49
  your entire purchase. 21%. That's a really solid
• 1:52
  discount. Just head over to pinsandaces .com.
• 1:55
  That's pinsandaces .com, code SNARFUL. Alright,
• 1:59
  let's get into it. We've got our internal snarfle
• 2:01
  tech article report the morning star ready to
• 2:03
  go three big stories today Yeah And our goal
• 2:06
  here is always is to break these down to connect
• 2:08
  the dots figure out what this tech news actually
• 2:10
  means for well for you Exactly. Let's start with
• 2:13
  this first one It's coming out of China and involves
• 2:16
  a company called deep seek making waves in AI
• 2:19
  efficiency deep seek They've they've popped up
• 2:23
  before haven't they made some noise? They definitely
• 2:25
  have. Their previous models apparently, quote,
• 2:28
  shocked Silicon Valley. Now they've released
• 2:30
  an experimental model, DeepSeek V3 .2 XBS. Experimental,
• 2:35
  OK. So maybe not fully baked yet. They're calling
• 2:38
  it an intermediate step towards whatever their
• 2:40
  next big thing is. But the key part, the tech
• 2:43
  nugget here, is something called DeepSeek Sparse
• 2:46
  Attention or DSA. DSA. OK, break that down for
• 2:49
  Sparse Attention. Right. So think about how these
• 2:52
  large language models usually work. They pay
• 2:55
  attention to all the words in a long piece of
• 2:58
  text to understand the context. It's computationally
• 3:01
  expensive. Like reading every single word in
• 3:03
  a giant manual. Exactly. DSA, the Sparse Attention,
• 3:07
  lets the AI be smarter about it. It focuses only
• 3:11
  on the most important bits of information, kind
• 3:13
  of skipping the less relevant parts. Like skimming
• 3:16
  the manual for the important diagrams and warnings.
• 3:18
  Precisely. And the result? DeepSeek claims it
• 3:21
  drastically cuts the computing cost and actually
• 3:23
  boosts performance, especially when dealing with
• 3:25
  really long sequences of text. Makes training
• 3:28
  these huge models way cheaper. Okay, cheaper
• 3:31
  training, better performance. That sounds good.
• 3:35
  But what's the catch? Or what's the impact? Well,
• 3:38
  here's the kicker. They're not just keeping those
• 3:40
  savings. They announced they're slashing their
• 3:42
  API prices by 50 % plus... 50 percent. Wow. OK,
• 3:47
  so that's not just a small adjustment. That's
• 3:49
  aggressive. It's basically firing a starting
• 3:51
  gun on a price war. It puts huge pressure on
• 3:54
  their Chinese rivals like Alibaba's Quinn, but
• 3:56
  also on the big international players. Think
• 3:59
  OpenAI and Thropic. If DeepSeek can offer similar,
• 4:04
  maybe even better performance for half the price.
• 4:07
  Yeah, that changes the whole game. And that's
• 4:09
  the connection for you listening. If the fundamental
• 4:11
  costs of running these AI tools, the ones powering
• 4:13
  your virtual assistants, your customer service
• 4:15
  chats, maybe even tools you use at work if that
• 4:18
  cost gets cut in half, then theoretically the
• 4:20
  cost of those services should eventually come
• 4:22
  down too. That's the idea. Efficiency improvements
• 4:24
  like this tend to trickle down. It makes powerful
• 4:26
  AI potentially much more accessible and affordable.
• 4:29
  Okay, very interesting. Efficiency driving down
• 4:31
  costs. Now our second story is about Well, the
• 4:36
  opposite of cheap, maybe. A massive amount of
• 4:38
  money changing hands. Yeah, a huge one. We're
• 4:40
  talking about the gaming giant Electronic Arts,
• 4:43
  EA. They're being taken private in a $55 billion
• 4:47
  buyout. $55 billion. That's enormous. It's being
• 4:51
  called the largest all -cash sponsor take private
• 4:54
  investment in history. All cash. It's a landmark
• 4:57
  deal for the gaming world, for sure. Who's behind
• 4:59
  such a massive deal that takes deep pockets?
• 5:01
  It's a group, a consortium, led by Saudi Arabia's
• 5:04
  public investment fund, the PIF. They already
• 5:07
  had a stake, about 9 .9%, and they're rolling
• 5:10
  that over. Also involved are Silver Lake and
• 5:13
  Jared Kushner's affinity partners. Okay, so major
• 5:15
  players and the deal itself. Shareholders get
• 5:18
  $210 a share. Yep, which is about 25 % premium
• 5:21
  over where the stock was. It's structured as
• 5:23
  $36 billion in equity and $20 billion in debt.
• 5:27
  They expect it to close in, let's see, Q1 of
• 5:29
  their 2027 financial year. And the current CEO,
• 5:32
  Andrew Wilson, he's staying. HQ stays in Redwood
• 5:34
  City. That's the plan. So on the surface, it
• 5:37
  might seem like stability. But going private
• 5:40
  changes things fundamentally. How so? Less pressure
• 5:43
  from the stock market every quarter. Exactly.
• 5:46
  Public companies often obsess over those quarterly
• 5:48
  earnings reports. Analysts think this deal is
• 5:51
  basically a bet that the games industry is now,
• 5:54
  quote, mature and predictable. Mature and predictable.
• 5:58
  Meaning they want the reliable money makers.
• 6:01
  Pretty much. Focusing on the huge franchises
• 6:04
  that churn out cash year after year. Think EA
• 6:07
  Sports FC, formerly FIFA, The Sims, Battlefield,
• 6:12
  big established IPs. Less focus on risky new
• 6:15
  ideas or smaller games. That's the potential
• 6:17
  downside, yeah. Private ownership means less
• 6:20
  pressure to gamble on the next big innovative
• 6:22
  hit that might take years to pay off. If ever,
• 6:25
  they might prioritize predictable profit streams.
• 6:27
  And EA has already been doing some restructuring,
• 6:29
  right? Layoffs shutting down some mobile games.
• 6:31
  Right. They shut down things like Kim Kardashian
• 6:33
  Hollywood, focusing more on their owned IP. This
• 6:35
  move kind of fits that pattern. So for gamers
• 6:38
  listening, what does this mean for like their
• 6:41
  favorite EA game? Does it mean? more reliable
• 6:44
  updates for Madden or FC, but maybe fewer totally
• 6:48
  new unexpected games from EA down the road. That's
• 6:51
  the big question, isn't it? It's potentially
• 6:52
  a trade -off. Stability for the big franchises
• 6:54
  versus maybe less risk -taking on groundbreaking
• 6:57
  new stuff. We'll have to see how it plays out.
• 6:59
  $55 billion question. Okay, that brings us to
• 7:02
  story number three. And speaking of staggering
• 7:05
  amounts of money, this one is about the sheer
• 7:08
  cost of building the infrastructure for this
• 7:10
  AI boom. Yeah, this puts the deep -seek efficiency
• 7:13
  gains in perspective. We're talking about Oracle
• 7:15
  and OpenAI here, and the numbers are just mind
• 7:18
  -boggling. Okay, hit me with it. So, OpenAI has
• 7:20
  reportedly signed a deal with Oracle, a contract
• 7:23
  for computing power. The price tag? $300 billion.
• 7:27
  $300. Wait, billion. With a B. Billion with a
• 7:29
  B. Over five years, starting in 2027, $300 billion
• 7:33
  just for the compute power from Oracle. I can
• 7:36
  barely wrap my head around that number. What
• 7:37
  does that even mean in physical terms? Get this.
• 7:40
  To deliver that much compute, the infrastructure
• 7:43
  required needs 4 .5 gigawatts of power. Continuously.
• 7:47
  4 .5 gigawatts. What's the comparison there?
• 7:50
  It's roughly the output of more than two Hoover
• 7:53
  dams. Or, put another way, enough electricity
• 7:56
  to power about four million homes in the US.
• 7:59
  Just for this one OpenAI contract? That's insane.
• 8:02
  We're literally talking about needing damn -level
• 8:05
  power generation just for AI. It's a staggering
• 8:08
  physical footprint. And think about the financial
• 8:10
  gamble here. OpenAI's current revenue is maybe,
• 8:12
  what, 10, 12 billion dollars a year? Right. This
• 8:15
  contract means they'd be paying Oracle 60 billion
• 8:18
  dollars a year starting in 2027. 60 billion!
• 8:21
  they're projecting huge losses already like 44
• 8:24
  billion dollars between 2023 and 2028 so they're
• 8:27
  bidding everything on absolutely massive exponential
• 8:29
  growth in ai demand and revenue to cover that
• 8:32
  cost completely it's an enormous bet and oracle's
• 8:35
  taking a huge gamble too right they have to actually
• 8:37
  build this out they'll likely take on significant
• 8:39
  debt to buy all the specialized tubes mostly
• 8:41
  from nvidia to fulfill this contract but oracle
• 8:43
  must see a massive upside if they're willing
• 8:45
  to do that Oh, for sure. Their contracted future
• 8:47
  revenue, what they call remaining performance
• 8:49
  obligations, or RPO, has apparently skyrocketed.
• 8:53
  It's up 359 % to $455 billion. $455 billion,
• 8:57
  okay. Yeah, driven by deals with basically the
• 9:00
  who's who of AI, OpenAI, sure, but also Elon
• 9:03
  Musk's... XAI, Meta, others, they're positioning
• 9:06
  themselves as the infrastructure provider for
• 9:09
  the AI age. They expect their cloud infrastructure
• 9:11
  revenue to jump 77 % in fiscal 26 alone. And
• 9:15
  the quiet giant in the background here is Nvidia,
• 9:18
  right? Making all the chips. Absolutely. Can't
• 9:19
  forget them. Their CEO, Jensen Huang, he's talking
• 9:22
  about total AI infrastructure spending reaching
• 9:24
  maybe three to four trillion dollars by the end
• 9:27
  of the decade. Brilliant. With a T. Trillion
• 9:28
  with a T. So that $300 billion Oracle deal, as
• 9:31
  mind blowing as it is, is just one piece of a
• 9:34
  multi -trillion dollar puzzle Huang envisions.
• 9:36
  Wow. So the connection for everyone listening,
• 9:39
  I mean, is this just the necessary cost of building
• 9:41
  the future or is it kind of spending fueled by
• 9:44
  debt and the projections setting us up for some
• 9:46
  kind of massive tech bubble? That's the trillion
• 9:48
  dollar question, isn't it? You've got DeepSeq
• 9:50
  pushing costs down with clever engineering on
• 9:53
  one side. Right. And on the other, you have this
• 9:55
  almost unimaginable infrastructure build out.
• 9:59
  costing fortunes and demanding incredible amounts
• 10:01
  of physical resources like power. The tension
• 10:04
  between those two is really shaping things right
• 10:06
  now. It really is. Okay. Fascinating stuff. Before
• 10:10
  we recap, let's quickly thank our other sponsor.
• 10:13
  Yes. We also want to thank Skoni. That's S -K
• 10:15
  -O -N -I. Skoni. They make premium golf shoes.
• 10:18
  If you're looking for that perfect combination
• 10:20
  of style, comfort, and performance on the course,
• 10:23
  you really should check out Skoni. And guess
• 10:25
  what? We've got a code for them too. Use promo
• 10:27
  code SNARFUL again. That's S -N -A -R -F -U -L
• 10:30
  SNARFUL at checkout on their website. Using that
• 10:33
  code gets you 15 % off your entire purchase.
• 10:36
  A great deal on some really top -notch golf shoes.
• 10:39
  That's scony .com code SNARFUL. All right. Let's
• 10:42
  quickly tie these three stories together from
• 10:44
  our Morningstar report. OK. First up, DeepSeek
• 10:46
  in China. Using that new DSA tech, DeepSeek sparse
• 10:50
  attention to drastically cut AI computing costs.
• 10:54
  Yeah, and launching a price war by slashing API
• 10:57
  costs by over 50 %? This should eventually make
• 11:00
  AI tools cheaper for everyone. Second story.
• 11:03
  Electronic Arts. The gaming giant going private
• 11:06
  in that record -breaking $55 billion all -cash
• 11:10
  deal. Right. Led by PIF and others. It signals
• 11:13
  a shift towards focusing on stable, predictable
• 11:16
  profits from big franchises, potentially at the
• 11:18
  expense of riskier, innovative new games. And
• 11:21
  third. the absolutely astronomical scale of the
• 11:24
  AI infrastructure build out. Illustrated by that
• 11:27
  $300 billion Oracle OpenAI deal needing 4 .5
• 11:30
  gigawatts of power more than two Hoover dams
• 11:32
  worth. And it's just part of a potential $4 trillion
• 11:35
  spend by 2030. So let's leave our listeners with
• 11:37
  a final thought on that infrastructure piece.
• 11:39
  Okay, so we're talking dam levels of power for
• 11:43
  just one AI contract and projections of trillions
• 11:48
  in spending overall. Forget the tech companies
• 11:50
  for a second. What existing public utility, think
• 11:55
  the power grid, maybe water systems needed for
• 11:58
  cooling data centers, even the fiber optic networks,
• 12:01
  which of those non -tech utilities is going to
• 12:03
  face the biggest strain, the biggest need for
• 12:05
  costly upgrades to actually support this multi
• 12:09
  -trillion dollar AI future? That's a great question.
• 12:12
  Where does the physical burden really fall when
• 12:15
  the digital demand gets this huge? something
• 12:18
  to think about. It connects right back to your
• 12:20
  bills, your services, everything. It really does.
• 12:22
  Alright, that's our time for today. We're here
• 12:23
  to help at Snarful, so reach out on our website
• 12:25
  if you have any more questions. Thanks for tuning
• 12:27
  in, and we'll catch you next time for more Connecting
• 12:29
  the Docs. Bye for now. Thanks for listening,
• 12:47
  and we'll catch you next time.