Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Tech Talks With Kinsoft

Tech Talks With Kinsoft artwork

Last Week in Tech – The AI Boom's Memory-and-Power Bill, Microsoft's 4,800 Cuts, and a Perfect-10 ColdFusion Flaw

Jul 12, 202623 minAudio

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Your Monday catch-up on the week in tech, with a security lens. This week the real AI story wasn't a model - it was the physical bill coming due. Samsung posts a record quarterly profit, up nearly nineteen-fold, on AI-memory demand; SK Hynix raises around 29 billion dollars in one of the biggest-ever US share listings; and Britain's National Grid tips 1.75 billion into a US power project built to feed a single Microsoft data centre. Microsoft cuts about 4,800 jobs and overhauls Xbox as it reallocates toward AI. Europe's top court makes Google's 4.1-billion-euro Android fine final. And on the security desk: a perfect-10 Adobe ColdFusion flaw (CVE-2026-48282) is exploited within two hours of disclosure and draws an emergency CISA patch order, Accenture confirms a breach after a criminal offers stolen data for sale, and INTERPOL's Operation First Light nets 5,811 arrests and 293 million dollars. Trying to work out what all this means for your hardware budget and your patch queue? Visit www.kinsoft.com.au to talk through your security and IT needs. Sources: Reuters; CNBC; Bloomberg; Samsung; TechCrunch; DataCenterDynamics; Adobe; CISA; BleepingComputer; SecurityWeek; Help Net Security; INTERPOL.

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

You know when you flip a light switch in your living room. You just you expect the light to turn on. Right. You definitely aren't thinking about the power grid. Exactly. You don't stand there visualizing the power plant 50 miles away or like the high voltage transmission lines cutting through the mountains. You just want the room to be bright. Yeah. We've been completely conditioned to expect the end result without ever really having to think about the mechanism behind it. And that expectation has just well, it's permeated how we interact with all modern technology. Oh, 100 percent. Yeah. The infrastructure is by design

completely. invisible to you. Which is exactly the illusion we are shattering today. Because we open our laptops, pull out our phones, type some complex query into an AI chat bot, and we just expect this highly synthesized magical answer in milliseconds. And you don't see the massive, humming, incredibly strained physical reality operating behind the scenes to make that happen. Right. So today we're taking a stack of really critical tech developments from mid -July 2026,

primarily anchored by this brilliant briefing called Tech Talks with Kinsoft. And we're going to actually look at the mechanism. Yeah, we aren't just running down a list of tech news today. No, not at all. The mission for this deep dive is to explore the hidden infrastructure of the technology you rely on every single day. We're looking at the physical constraints, the ruthless corporate maneuvers, the regulatory crackdowns, and honestly, the terrifying speed of modern security threats. Because, I mean, if you are managing any kind of technology stack or business

strategy right now, understanding that hidden infrastructure is really the only way you can accurately map out your own risk. And your budget, for that matter. So true. The reality of tech is just far more physical and far more vulnerable than a seamless user interface implies. Absolutely. So let's start at the literal foundation. Before any software can execute, before any large language model can write a line of code or generate an image, it fundamentally requires hardware and electricity. Yeah, you can't code your way out of needing a power source. Exactly. And based on three separate financial announcements this

week, an enormous physical bill is coming due for the AI boom. Oh, it's an infrastructure build out on a scale we really haven't seen in decades. The numbers are almost difficult to process. They really are. Let's look at Samsung's preview of their second quarter results from July 7th. They're projecting an operating profit of roughly 89 trillion Korean won. Which is what, about 58 billion U .S. dollars? Yeah, almost 60 billion. And that profit is up nearly 19 -fold from a year earlier. 19 times. That's just wild. It's

the highest quarterly profit in their entire corporate history. And this isn't from selling more TVs or smartphones. It is almost entirely driven by the explosive demand for specialized high -bandwidth memory chips, or HBM for AI data centers. And then a few days later, SK Hynix debuts on the U .S. NASDAQ exchange, pulling in around $29 billion specifically to fund new chip fabrication and packaging plans. Yeah, that SK Hynix debut is one of the largest listings the U .S. market has ever seen. Massive. It is.

And just to put some technical context behind why these specific chips are generating $58 billion in profit for Samsung, it really comes down to how AI actually computes. Okay, wait on us. So traditional computing is largely sequential. But training an AI model requires parallel processing, meaning you have to shuffle massive, unprecedented data sets back and forth between the processor and the memory simultaneously. So a standard chip just can't keep up. Exactly. Standard memory chips create a traffic jam. High bandwidth memory

is essentially widening a two -lane highway into a hundred -lane superhighway. AI data centers literally cannot function without it. Wow. And then there's the power required to actually run that superhighway. Britain's National Grid announced a $1 .75 billion investment into a U .S. power developer to build a 2 .5 gigawatt facility. That's a lot of juice. And here's a kicker. Its entire purpose is to feed a single Microsoft data center campus in Texas. OK, let's unpack this. A typical nuclear power plant generates

about one gigawatt of power. So we are talking about dedicating two and a half nuclear power plants worth of electricity to a single cluster of buildings. It's like everyone is admiring a shiny new sports car, the AI chatbots. But nobody is noticing that the industry is fundamentally running out of the asphalt and gasoline required to actually drive them. That's a great way to put it. The constraints have entirely shifted for the last 20 years. The tech industry's bottleneck was software engineering, right? Like, how fast can you write the code? Today, the bottleneck

is down in the physical layer. Can you secure the silicon and can you secure the megawatts? So let's bring this down to the ground level for you, the listener. If I'm an IT director mapping out my three -year hardware budget or a CTO planning a major infrastructure overhaul, why does a giant Texas data center affect me? How does this massive supply suction at the top actually hit my spreadsheets? Well, it hits your spreadsheets in two very distinct ways, cost and time. These massive data center builders, you know, your Microsoft, Google's, Amazon's,

they are quite literally hoovering up the global supply of raw materials. Mostly memory and silicon. Exactly. When buyers at that scale are absorbing all the inventory, the trickle -down effect for enterprise hardware is severe. So regular businesses pay the price. Yeah, you need to adjust your forecasting immediately. Expect memory to be significantly pricier across all your server and endpoint purchases, and expect lead times for network equipment to stretch out. Oh, that makes sense. Yeah, if you used to refresh your hardware every four years, you honestly need

to start the procurement process at year three now. So the physical constraints of AI are going to tax the budgets of companies that aren't even building AI. Precisely. And it also changes how you evaluate vendor promises. When an enterprise software vendor tells you a revolutionary new AI feature is coming, like next quarter, you have to realize that the honest barrier to that release isn't necessarily their software team anymore. It's whether they have the server space. Exactly. It's whether their cloud provider has the physical server capacity to host it. So when

product roadmaps slip this year, the physical limit of the grid is very likely the hidden reason why. has always preached this gospel of infinite digital scaling. And now they were just violently hitting the hard limits of the physical world. They really are. But this pivot to AI isn't just sucking up electricity and silicon. It is actively cannibalizing other departments within these very same tech giants. We are moving from the physical reallocation of resources to the human reallocation. Which brings a completely different

kind of instability to the enterprise landscape, particularly regarding the vendors that almost every business relies on. Yeah, let's look at Microsoft. On July 6th, they cut around 4 ,800 jobs, which is a bit over 2 % of their workforce. The gaming division took the absolute heaviest blow. Xbox shed roughly 20 % of its staff, and they sold off or spun off several studios. That's a huge cut for one division. Yeah, the internal memo from Xbox leadership was incredibly blunt. They stated, our business today is not healthy.

But they didn't just cut gaming. Microsoft used this round of layoffs to fundamentally reshape its commercial sales and consulting organizations. Here's where it gets really interesting. Microsoft just posted quarter after quarter of record -breaking revenue. Right, they're making money hand over fist. Exactly. This is not a company in distress trying to avoid bankruptcy. It's like aggressively pruning a perfectly healthy, mature oak tree just so one specific branch AI and cloud can grow to a massive, monstrous size. What's fascinating

here is that you're watching a deliberate, ruthless reallocation of capital in real time. In a high interest rate environment, even a giant like Microsoft doesn't have infinite cash to throw at every project. Right. So they are looking at their portfolio, pulling money, attention and highly skilled personnel out of slower growing legacy areas and pouring absolutely everything into AI infrastructure. They are reshaping the entire corporate organism to win one specific arms race. But if I am an enterprise customer. relying on Microsoft's commercial consulting arm to architect my cloud migration, what is

the ripple effect of that internal pruning on my business? The ripple effect is a sudden loss of institutional knowledge and relationship continuity. Enterprise technology at scale is well, it's heavily reliant on relationships. Yeah, the people you talk to. Exactly. When a vendor fundamentally restructures its commercial sales and support teams to change a new strategic priority, the trusted account manager who understands your specific compliance needs might just vanish. The Tier 3 support structure you're used to navigating

can be completely rewritten overnight. Wow, so you can't just assume the organizational chart of your vendor is a... static resource. Exactly. You need to map out your dependencies now, know exactly who your critical contacts are, and have a clear documented escalation path for how to get enterprise support if your entire account team is suddenly reallocated to an AI initiative, or just manage out of the company entirely. Now, with tech giants ruthlessly consolidating power, money, and focus to dominate this new era of

cloud computing, you'd think they're entirely unstoppable. But that kind of aggressive consolidation makes governments incredibly nervous. Oh, definitely. Regulators are pushing back against monopolistic behaviors and they are finally proving they can draw blood. Yeah. The regulatory landscape, particularly in the European Union, has crossed a very significant threshold this month. Right. So on July 2nd, the European Union's highest court threw out Google's final appeal against a massive antitrust fine. We are talking about 4 .1 billion euros.

So roughly 4 .7 billion U .S. dollars. Huge number. The core issue was how Google bundled its search app and Chrome browser onto Android phones, essentially forcing device manufacturers to pre -install them if they wanted access to the Google Play Store. Which is a textbook example of leveraging dominance in one market, the Android operating system, to choke out competition in another market like search and web browsing. Just by making it the default. Exactly. By forcing manufacturers to set Chrome and Google search as the default

out of the box, they capitalized on user inertia. most users simply will not change the default browser on a new phone i know i don't right so that effectively locked competing search engines out of the mobile ecosystem before the user even turned the device on okay i understand the mechanics of the monopoly argument there but i have to push back on the impact 4 .7 billion dollars is a staggering number to you and me But for a titan like Alphabet, a company that generates tens of billions of dollars in profit every single

quarter, isn't this ultimately just an expensive parking ticket? Like, hasn't this traditionally been the acceptable cost of doing business for big tech? You break a rule, you capture a market, you pay a fine a decade later, and you keep your dominance. So that has historically been the playbook, yeah. But this ruling signals that the parking ticket era is over. The dollar amount grabs the headlines, but the legal finality is what actually matters here. This specific case has been dragging through the appeals process since 2018. Google employed some of the most

expensive legal teams on earth to fight this, and they lost their final appeal. The courts are firmly... undeniably backing the regulators. So the strategy of tying up regulators in infinite litigation is failing. It is failing and the fines are sticking. For you, the listener, this fundamentally changes the risk calculus. If your organization operates in Europe or if you're developing a product strategy that relies on aggressively bundling software to force user adoption, you have to recognize that the regulatory bodies now have real unavoidable teeth. You can't

just stall them forever. Exactly. You can no longer assume that your legal department can just litigate away anti -competitive behavior. The blueprint for prosecuting big tech in the EU is now finalized and legally binding. So while these mega corporations are fighting regulators over market control in the courts, the actual day -to -day software infrastructure that underpins all of these companies is constantly crumbling under our feet. Yeah, the velocity of the security landscape right now is frankly terrifying. It really is. Yeah. Let's talk about Adobe ColdFusion.

On June 30th, Adobe patched a batch of flaws in this web application platform. One of them, cataloged as CVE -2026 -48282, scored a perfect 10 out of 10 on the CVSS severity scale. Which is as bad as it gets. Right. For context, a 10 out of 10 usually means an attacker requires zero authentication, no username, no password, to remotely execute their own malicious code on your server from across the Internet. And according to the Kinsoft briefing, security researchers observed attackers actively exploiting this specific

flaw within two hours of Adobe publicly disclosing it. Two hours. That is just two hours. The U .S. cybersecurity agency CISA immediately added it to their known exploited vulnerabilities catalog and ordered all federal agencies to patch it. A 10 out of 10 severity weaponized and actively exploited in two hours. That is the brutal reality of modern cyber warfare. So what does this all mean? Because I am looking at that two hour window. And the operational math just doesn't work for most IT departments. It's like a burglar finding a highly detailed blueprint to your house and

then robbing it before the ink on the blueprint is even dry. That's a really good analogy. If an exploit is live in two hours, the traditional enterprise mindset of we'll test the patch and deploy it during the maintenance window next weekend. is literally a death sentence for your network. Oh, it completely invalidates legacy patching schedules. And what makes cold fusion particularly dangerous is that it's often legacy middleware. Meaning it's kind of forgotten about. Exactly. It's the connective tissue lurking behind older, half -forgotten web applications that

no one on the current IT staff actually built. If you run software like that anywhere on your perimeter, you have to automate your patching. If you genuinely cannot patch a 10 out of 10 vulnerability within hours of disclosure, Your only other option is to sever that server from the public Internet until you can. The grace period between disclosure and disaster is gone. And it's not just older middleware getting hit. The giants themselves are bleeding. The briefing notes that on July 7th, the massive global consulting and IT services firm Accenture confirmed they

suffered a security incident. A cyber criminal. on a dark web forum claimed to be selling a massive cache of stolen Accenture data. Yeah, I saw that. Now, Accenture stated the matter was isolated, that they immediately remediated the source, and that there was no impact on their operations or their clients' systems. Okay, if we connect this to the bigger picture, we need to model some critical thinking regarding how we consume threat intelligence. Right, because there's always a lot of noise. Yes. When dealing with a high

-profile breach notification, you have to surgically separate the confirmed facts from the threat actor's claims. I see the distinction, because the cybercriminal claimed they had extracted 35 gigabytes of data, specifically highlighting that they had internal source code and a batch of cloud access keys. Precisely. The confirmed fact is that a perimeter breach occurred at Accenture. The unverified claim is the 35 gigabytes of source code and keys. The journalists who broke the story couldn't verify the full scope of that data dump. So they might just be bluffing to

drive up the price. Exactly. And this is the crucial takeaway for the listener. The core lesson holds, regardless of whether it was 35 gigabytes or 35 megabytes. Okay, let's pause and clarify the technical threat there. What? When we say cloud access keys, we are basically talking about machine -to -machine master passwords, right? Pretty much, yeah. These are tokens that allow automated scripts or software to bypass human multi -factor authentication and go directly to cloud infrastructure. That's exactly it. They are the keys to the kingdom. And if a global

giant with immense world -class security resources like Accenture can potentially leak source code and access tokens, your third -party suppliers absolutely can too. Wow. Yeah. It is the ultimate supply chain risk. Think of it like giving a valet the keys to your corporate vehicle. You might have the best car alarm in the world, but if the valet company's unmonitored lockbox gets pried open, your car is still getting stolen. Man. So if I am a CTO, how do I defend my architecture against a breach that happens on my supplier's

network? I can't patch Accenture servers. You know, you can't. You build defense in depth on your own side of the wall. The defensive moves are entirely unglamorous, but they are essential. First, you mandate strict automated rotation of your cloud keys and access tokens. So they expire quickly. Right. If a token gets stolen in a third -party supply chain breach, you want to ensure it expires before the attacker can figure out what it unlocks. Second, you lock down your code repositories. Implement strict identity access management so that only specific...

That makes a lot of sense. And finally, review your vendor contracts today. Ask your major IT suppliers point blank what their breach notification protocols are. If their lockbox gets pried open, how many hours will pass before they notify you? You must secure those answers before an incident happens, not while you are reading about it on a security blog. Okay, so assume the supply chain breach is inevitable and aggressively limit the shelf life of what can be stolen. is highly actionable. Now, we've talked about billion dollar physical

power constraints, 10 out of 10 software vulnerabilities and complex supply chain breaches. But our final story highlights that the most consistently successful attack vector on the planet doesn't rely on cutting edge zero day code at all. No, it doesn't. It relies entirely on human psychology. Exactly. The human mind remains the most exposed, unauthenticated attack surface in any enterprise. On July 9th, Interpol announced the results of an operation called First Light 2026. This was huge. It was a massive, highly coordinated law enforcement

crackdown spanning 97 countries. They made over 1 ,500 arrests globally and intercepted $293 million in illicit stolen money. Just an incredible haul. And what kind of high -tech digital wizardry were these syndicates using? It wasn't remote code execution. It was everyday scams, phishing, romance cons, investment fraud, and overwhelmingly business email compromise. or BEC. Yeah, BEC is a nightmare. This is where a crook impersonates a CEO or a trusted vendor, creates a false sense

of urgency, and simply asks an employee to route a payment to a fake invoice. $293 million intercepted globally, fundamentally sourced from variations of criminals just politely or urgently asking people to hand over their money. Which brings up a fascinating philosophical contradiction to me. With all this incredibly complex technology we've explored today, the AI data centers, the multi -cloud architectures, the sophisticated software flaws, why is rudimentary social engineering still the undisputed workhorse of global fraud? Are we, as humans, just the unpatchable bug in

the system? This raises an important question about how threat actors actually operate. And the answer is rooted in pure economics. Hacking is a business. Right. They want the best ROI. Exactly. Tricking a stressed, multitasking executive into paying a fraudulent invoice is fundamentally easier, cheaper, and yields a higher return on investment than spending six months trying to brute force a corporate firewall. That makes total sense. And these BEC attacks have evolved too. They aren't just spoofing email addresses anymore. They are compromising legitimate internal

email threads, watching the cadence of the conversation and inserting the fake invoice at the exact moment a payment is expected. inherently want to be helpful. They want to resolve urgent issues quickly and attackers ruthlessly weaponize that empathy and urgency. Wow. So if you can't push a software update to the human brain, what is the defense? How do you secure an organization against its own empathy? You don't patch the human, you patch the process. The single most effective control against business email compromise is incredibly boring, but it works flawlessly if enforced.

What is it? You institute an unbreakable rule. If a payment detail, a routing number, or a bank account changes on an invoice, you must verify that change through a secondary out -of -band channel before moving a single cent. A secondary out -of -band channel. Meaning, under no circumstances do you just hit reply to the email asking for the money. Never. You pick up the phone and call a number you already have saved on file for that supplier. You don't call the new phone number listed in the suspicious email signature. You

use a known, trusted, pre -existing channel to verbally confirm the change. Oh, wow. It's that simple. It adds exactly two minutes of friction to the account's payable process, and it stops millions of dollars in corporate fraud dead in its tracks. It is remarkably sobering how the simplest administrative processes are often the only effective firewall we have. Well, we have covered some incredibly dense territory today on this deep dive. We started with the massive gigawatt level physical constraints of the AI boom. The memory and megawatts fundamentally

required just to keep the lights on. The literal founding. Exactly. We explored Microsoft's ruthless human capital reallocation to feed that very same infrastructure. We unpacked how the EU's antitrust regulators are finally proving their fines have real unavoidable teeth. We looked at the terrifying two -hour exploitation window of software vulnerabilities and the reality of third -party supply chain breaches. It's been a heavy list today. It really has. And finally, we ended on the enduring, unpatchable vulnerability of human psychology and the economics of social

engineering. Because the infrastructure of technology is not a clean digital cloud. It is a highly complex physical machine, deeply intertwined with raw materials, corporate strategy, regulatory friction, and human behavior. And understanding that messy, hidden infrastructure, staying critically informed about the mechanisms behind the headlines, is truly the best defense you can have against a rapidly shifting tech landscape. Thank you for joining us today, for going past the shiny interfaces, and for taking this journey with us. We've spent a lot of time today talking about

two distinct problems, how AI development is currently constrained by massive physical power needs, and separately, how cyber attacks are moving with terrifying speed, weaponizing flaws in just two hours. But I want to leave you with a final thought to mull over. What happens when those two isolated forces inevitably collide? Ooh, what do you mean? What happens when generative AI is fully unleashed by threat actors to optimize and automate those rapid cyber attacks? and they point those automated attacks directly at the strained physical power grids and data centers

that AI relies on to exist. It is a technological snake eating its own tail, and the fallout of that collision is something every business leader needs to be preparing for right now.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Iris Ohyama - A Suspicious File in the Cloud, 7.7 Million Customers at Risk, and an Online Store That Went Dark

    Nov 19, 202620 minAudio

    Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Japanese home goods company Iris Ohyama (online store operator Iris Plaza) found a suspicious file in an external...

    Credit the people on this episode

  2. Who Gives A Crap - A Marketing Supplier Breached, Six Days to Tell the Client, and Customer Data That Didn't Need to Be There

    Nov 17, 202619 minAudio

    Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Australian online retailer Who Gives A Crap told customers on 7 November 2026 that their information was accessed...

    Credit the people on this episode

  3. Last Week in Tech - An Outlook Zero-Day That Leaks Your Password Hash, Chrome's Eighth Exploited Bug, and Nvidia Without Jensen Huang

    Nov 15, 202621 minAudio

    Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Last week in tech, 7 to 14 November 2026. Microsoft's November Patch Tuesday (10 November) fixes well over a...

    Credit the people on this episode

  4. Chainalysis - From "Non-Sensitive" to "All Users": How a Breach Notice Changed in Five Days

    Nov 12, 202619 minAudio

    Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! Chainalysis disclosed on 30 October 2026 that an external actor using "offensive agentic techniques" accessed...

    Credit the people on this episode

  5. NSW Health - A Dental Voucher Website, Four Gigabytes Taken, and the Providers Whose Details Were Exposed

    Nov 10, 202623 minAudio

    Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! NSW Health reported unauthorised access to its Oral Health Fee For Service Scheme website on 29 October 2026;...

    Credit the people on this episode