Latest / Connecting the Dots / AI’s Billion-Dollar Battles, Intel’s Panther Lake Chips, and Google’s Enterprise AI Push

Transcript

AI’s Billion-Dollar Battles, Intel’s Panther Lake Chips, and Google’s Enterprise AI Push · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarfle 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. Hello and welcome to Connecting the Dots.
• 0:17
  I'm Alex. And I'm Morgan. Thanks for tuning in.
• 0:20
  It is Thursday, October 9th, 2025. Here in Sacramento,
• 0:25
  it's looking like a beautiful day, becoming sunny
• 0:28
  with a high around 75. Nice. Nationally, though,
• 0:32
  things are a bit more active. We've got serious
• 0:35
  tropical moisture from what was Priscilla causing
• 0:38
  some, well, pretty widespread heavy rain and
• 0:41
  flash flood risks. Right, especially across the
• 0:44
  Southwest and Four Corners region. Burn scars
• 0:47
  in urban areas are particularly vulnerable. Yeah,
• 0:49
  definitely something to keep an eye on there.
• 0:51
  And also some widespread storms, maybe isolated
• 0:54
  flash flooding, possible down in Central and
• 0:56
  South Florida, too. OK, quick look at the markets
• 0:58
  before we jump in. Yesterday, things were a little
• 1:01
  mixed. The Dow Jones finished slightly down,
• 1:03
  just 0 .10%, about 47 points. But the S &P 500
• 1:07
  actually showed some strength, up almost half
• 1:09
  a percent, adding about 30 points. And Bitcoin,
• 1:12
  pretty much flat, down just a tiny, tiny fraction,
• 1:15
  0 .02%. So some stability there, which kind of
• 1:17
  feeds into our first story, doesn't it, with
• 1:19
  all this AI investment happening? It really does.
• 1:22
  But first, just a quick word about one of our
• 1:24
  sponsors. Absolutely. This show is brought to
• 1:26
  you by Pins and Aces. They make fantastic premium
• 1:28
  golf apparel and accessories. Really cool stuff.
• 1:32
  Great gear. And if you're listening, you can
• 1:33
  use the promo code SNARFUL. That's S -N -A -R
• 1:36
  -F -U -L at checkout. Yep. Use that code SNARFUL
• 1:39
  and you get 21 % off your entire purchase. Just
• 1:42
  head over to pinsandaces .com. Okay, so let's
• 1:45
  get into our Morningstar report. First up, this
• 1:47
  global race for open source AI, it feels like
• 1:50
  everyone's reacting to China's deep -seek model,
• 1:53
  you know, powerful, free. Totally. And now we're
• 1:57
  seeing major contenders emerge trying to offer
• 1:59
  alternatives. The big news is from the U .S.
• 2:02
  player, Reflection AI. Yeah, Reflection AI. Barely
• 2:06
  a year old right and they just announced two
• 2:08
  billion dollars in new funding and video is leading
• 2:11
  that round two billion That's staggering their
• 2:13
  evaluation just shot up from what around five
• 2:16
  hundred forty five million dollars back in March
• 2:18
  eight billion now eight billion dollars is just
• 2:20
  Wow So clearly the goal is to create that Western
• 2:23
  open source counterweight to deep seek exactly
• 2:25
  and it shows this well this investor fervor for
• 2:28
  AI isn't slowing down I mean look at the numbers
• 2:32
  foundation model companies worldwide. They've
• 2:35
  raised almost 72 billion dollars this year. That's
• 2:39
  more than double the 34 .9 billion dollars raised
• 2:42
  in all of last year. It really highlights how
• 2:44
  nations are viewing AI capability now. It's almost
• 2:47
  like a strategic national security kind of asset.
• 2:50
  It absolutely is. Which brings us nicely to the
• 2:52
  other big player that emerged K2 Think from the
• 2:55
  UAE. Released by this Emirati lab, the Institute
• 2:58
  of Foundation Models or IFM. And they're making
• 3:01
  some pretty bold claims, aren't they? They are.
• 3:03
  They're saying it rivals OpenAI's ChatGPT, even
• 3:06
  DeepSeq, on standard tests. But the really interesting
• 3:09
  part, the kind of, aha, moment. Is the efficiency.
• 3:12
  Exactly. It gets this performance using only
• 3:14
  32 billion parameters. 32 billion when the DeepSeq
• 3:17
  R1 is, like, 671 hours. Just precisely. It's
• 3:20
  apparently outperforming reasoning models that
• 3:22
  are 20 times larger. That points towards much
• 3:24
  lower costs, easier deployment. Maybe the race
• 3:27
  for just sheer size is shifting. OK. But is there?
• 3:29
  They're a catch. I mean, it's focused pretty
• 3:31
  narrowly, right? Math, coding, scientific research.
• 3:36
  That's the strategy. Hyper focusing allows them
• 3:39
  to hit those benchmarks and directly supports
• 3:41
  the UAE's big national AI plan, AI 2031. So it's
• 3:46
  less about a general purpose model and more about
• 3:48
  targeted capability. For now, yes. And it shows
• 3:50
  how the field is changing. You've got the UAE,
• 3:53
  Saudi Arabia, France with Mistral, Singapore.
• 3:55
  They're all seriously challenging that traditional
• 3:58
  US -China dominance in AI. The whole market's
• 4:01
  diversifying and efficiency doing more with less.
• 4:05
  That seems to be a huge selling point now. Which
• 4:07
  impacts future costs, access, everything, really.
• 4:11
  And of course, all this advanced AI needs advanced
• 4:13
  hardware to run on, especially locally. Which
• 4:16
  leads us perfectly into our next story, Intel's
• 4:18
  big reveal. Panther Lake. Right, the Core Ultra
• 4:21
  Series 3. This is their next -generation mobile
• 4:23
  chip, expected late 2025 or early 2026, and Intel's
• 4:27
  making a big promise here. They're basically
• 4:29
  saying you don't have to choose between high
• 4:30
  performance and high efficiency anymore, trying
• 4:32
  to merge the best of both worlds, like their
• 4:34
  previous Liner Lake and Air Lakeage chips. And
• 4:36
  the reason they think they can pull this off
• 4:38
  is the underlying tech. This is their first client
• 4:41
  ship using the 18A process. Which is essentially
• 4:44
  their two nanometer equivalent. Really cutting
• 4:46
  edge stuff. Yeah, and it relies on two key innovations.
• 4:50
  First, there's ribbon FET. Think of it like fully
• 4:53
  wrapping the transistor gate to stop electricity
• 4:55
  from leaking out. Okay, better efficiency there.
• 4:57
  And the second one. Power Via. This delivers
• 5:00
  power to the chip through the back, not the front.
• 5:03
  It allows for, well, greater density and, again,
• 5:06
  better efficiency. That's the engineering behind
• 5:08
  the performance claims. And those claims are
• 5:10
  pretty bold for laptop users. Up to 50 % more
• 5:13
  multi -threaded CPU and GPU performance compared
• 5:16
  to Lunar Lake. While consuming up to 10 % less
• 5:19
  power overall. So you should see faster speeds
• 5:22
  and genuinely better battery life in real -world
• 5:24
  use. That's the dream combo for anyone upgrading
• 5:27
  their laptop in, say, 2026. Definitely. And these
• 5:31
  chips will come in different flavors too, catering
• 5:33
  to different needs. You know, an 8 -core version,
• 5:35
  a 16 -core with decent graphics, and a 16 -core
• 5:38
  with much beefier graphics. Covering everything
• 5:41
  from basic productivity to maybe even some decent
• 5:44
  gaming on the go. Exactly. And speaking of gaming,
• 5:47
  the new XC3 graphics are supposed to offer over
• 5:50
  50 % more GPU power. Plus, they're adding multi
• 5:54
  -frame generation. Ah yes, the AI -generated
• 5:56
  frames. This is where the chip essentially inserts
• 5:59
  made -up frames between the real ones to make
• 6:02
  motion look smoother and boost frame rates. It
• 6:05
  can dramatically improve how games feel, yeah.
• 6:07
  But it's definitely a bit controversial. People
• 6:09
  call them fake frames. Right, because is it really
• 6:12
  faster performance? Or is it just clever visual
• 6:15
  trickery? Does it add latency? That's the debate.
• 6:18
  For serious competitive gamers, maybe not ideal,
• 6:20
  but for the average person playing a game on
• 6:22
  their laptop, it'll likely just look and feel
• 6:25
  smoother. OK, fair enough. What about the AI
• 6:27
  side, the MPU? Yep. It includes the MPU -5, rated
• 6:30
  at 50 TOPS, that's trillion operations per second.
• 6:34
  That handles those on -device AI tasks we keep
• 6:36
  talking about, making the whole AI PC concept
• 6:39
  more real. So processing AI stuff locally, faster,
• 6:43
  without always needing the cloud. Got it. There's
• 6:46
  one more feature though that caught my eye, Bluetooth
• 6:48
  Orcast. Oh yeah, this is actually really cool.
• 6:50
  Orcast lets multiple people connect their own
• 6:53
  Bluetooth headphones to the same audio source.
• 6:57
  So like... watching a movie on a plane with a
• 6:59
  friend and you both get good audio through your
• 7:02
  own preferred earbuds exactly no more sharing
• 7:04
  one headphone or using those cheap airline ones
• 7:07
  it's a small thing but a huge quality of life
• 7:10
  improvement for shared listening okay i like
• 7:12
  that so panther lake looks like a really crucial
• 7:14
  chip for intel pushing performance, efficiency,
• 7:17
  AI, and even some neat features like Oracast.
• 7:20
  It really sets the stage for the next wave of
• 7:23
  laptops and connecting back all this powerful,
• 7:25
  efficient hardware and the AI models running
• 7:27
  on it. They're getting packaged up for the workplace.
• 7:30
  Which brings us to our third story, Google Cloud
• 7:32
  Launching Gemini Enterprise. This is Google taking
• 7:35
  direct aim at Microsoft Copilot and OpenAI Enterprise.
• 7:39
  They're moving into that workplace automation
• 7:41
  space in a big way. And the focus here seems
• 7:43
  to be on agentic AI. Should probably unpack what
• 7:47
  that means. Right. So think beyond just a simple
• 7:50
  chat bot answering one question. Agentic AI is
• 7:54
  more like. like an invisible assistant that can
• 7:56
  handle complex multi -step tasks across different
• 8:00
  software. So it could say pull data from one
• 8:03
  system, draft a report based on it, maybe check
• 8:06
  it against some compliance rules in another document.
• 8:09
  And then automatically send that summary to your
• 8:11
  manager via, I don't know, Microsoft Teams or
• 8:14
  Slack for approval. It's about automating entire
• 8:17
  workflows, not just single actions. Okay, that's
• 8:19
  the promise of Gemini Enterprise and the price.
• 8:22
  Starts around $30 per user per month. Yeah, $30
• 8:25
  for the standard or plus tiers, a bit less, $21
• 8:28
  a month for their business plan. Google sees
• 8:30
  a big opportunity here. And are companies actually
• 8:33
  using these agents yet? They highlighted Virgin
• 8:35
  Voyages as an early adopter. Apparently they've
• 8:38
  rolled out over 50 specialized agents. One called
• 8:41
  Email Ellie handles personalized marketing emails.
• 8:44
  Email Ellie, okay. Yeah. And they're seeing positive
• 8:46
  results. They reported positive financial results,
• 8:48
  yeah. suggesting these automated workflows can
• 8:51
  actually boost efficiency and maybe even revenue.
• 8:54
  Now Google Cloud is growing, but Microsoft's
• 8:56
  cloud revenue is still much larger. What's Google's
• 9:00
  edge here? How are they competing? Integration
• 9:02
  seems key. Obviously, it connects tightly with
• 9:05
  Google Workspace. But crucially, they're also
• 9:08
  building connectors for competitors like Microsoft
• 9:10
  365, Salesforce, SAP. So they're trying to be
• 9:14
  the central hub, even if a company uses multiple
• 9:16
  platforms. Exactly. And providing a central governance
• 9:20
  framework, a way to manage and audit all these
• 9:22
  AI agents, no matter where the data lives. That's
• 9:25
  a big selling point for large companies worried
• 9:27
  about control and compliance. OK, but we have
• 9:30
  to talk about the name. Gemini it's a chatbot.
• 9:33
  It's the underlying AI model and now it's this
• 9:35
  big enterprise platform to isn't that confusing
• 9:37
  Oh, absolutely. It's causing definite friction
• 9:39
  I saw a Gartner analyst quoted basically saying
• 9:42
  this kind of naming confusion makes companies
• 9:44
  hesitant Yeah, like which Gemini are we even
• 9:46
  talking about now, right? The analysts suggested
• 9:49
  vendors are often just reacting, naming things
• 9:52
  as they build them, almost faster than they can
• 9:54
  strategize. And that uncertainty isn't great
• 9:57
  when you're asking businesses to make huge investments
• 10:00
  in a platform. A fair point. It's a major move
• 10:03
  for Google Cloud trying to monetize their AI
• 10:06
  and bring this complex automation to potentially
• 10:08
  millions of workers. Yeah. But maybe they need
• 10:11
  a clear branding strategy. Could definitely help.
• 10:13
  It's a space to watch closely. OK, really fascinating
• 10:16
  developments across the board. today. Before
• 10:18
  we wrap up, one more thank you to our other sponsor.
• 10:21
  Yes. We're also supported by Skoni, the premium
• 10:23
  golf shoe store. If you need shoes that perform
• 10:26
  great on the course, but also look sharp off
• 10:27
  it, check them out. And again, for you, our listener,
• 10:30
  use that same promo code SNARFUL, S -N -A -R
• 10:33
  -F -U -L, at checkout. Use SNARFUL and you'll
• 10:36
  get 15 % off your entire purchase at skoni .com.
• 10:39
  Great deal. Okay, so let's quickly recap the
• 10:42
  big themes today. First, that global AI race
• 10:45
  is really heating up and diversifying. We saw
• 10:48
  Reflection AI get that huge funding boost, but
• 10:50
  also the UAE's K2 think, showing that super efficient
• 10:54
  smaller models can compete. Yeah, maybe efficiency
• 10:57
  is becoming just as important as sheer scale
• 10:59
  now. Then we dove into the hardware enabling
• 11:01
  all this, Intel's Panther Lake chip. Right. Promising
• 11:04
  that blend of high performance and high efficiency
• 11:06
  for laptops coming in 2026 thanks to tech like
• 11:10
  the 18a process ribbon FET power via and that
• 11:13
  mpu5 for AI tasks plus or a cast. Don't forget
• 11:17
  our a cast And finally, we saw how this tech
• 11:20
  stack translates to the enterprise with Google
• 11:23
  Cloud launching Gemini Enterprise. Trying to
• 11:25
  automate those complex workflows using aogenic
• 11:27
  AI, connecting across different systems, and
• 11:30
  really challenging Microsoft in the workplace
• 11:31
  productivity space. A lot of dots connecting
• 11:33
  there. Definitely. So the final thought we'll
• 11:36
  leave you with today. If these smaller, incredibly
• 11:38
  efficient models like K2 Think can genuinely
• 11:41
  rival the performance of giant models 20 times
• 11:44
  their size, what does that mean for the future?
• 11:47
  of those massive, energy -hungry, large language
• 11:50
  models that currently dominate. Is the AI size
• 11:53
  race actually kind of over? That is definitely
• 11:57
  something to think about. A really interesting
• 11:59
  question for where AI goes next. Indeed. Well,
• 12:02
  that's all the time we have for today. Thanks
• 12:04
  for joining us. We're here to help at Snarful,
• 12:06
  so reach out on our website if you have any more
• 12:07
  questions. That's a wrap on today's top stories.
• 12:14
  Remember, tech doesn't have to be complicated,
• 12:16
  and we're here to help connect the dots. If you've
• 12:19
  got questions or want to learn more, visit us
• 12:21
  at snarful .com. Thanks for listening, and we'll
• 12:25
  catch you next time.