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 Capex Reckoning, AMD's Anthropic Deal, and WordPress Under Attack

Jul 26, 202619 minAudio
ShareXLinkedInFacebookWhatsAppEmail

Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode! The market finally asks AI to show its receipts. This week: Alphabet and Tesla beat on revenue but get hammered for record capex - the Nasdaq's worst day since April - while Intel posts its best growth since 2011; AMD and Anthropic strike a 2-gigawatt GPU deal with up to US$5B in equity; the "wp2shell" WordPress core flaws go from patch to mass exploitation in 48 hours; a 7,600-repository GitHub campaign booby-traps fake AI agent integrations - and AI coding assistants recommend them to victims; Anubis claims the Fairlife ransomware attack (unverified); and DeepSeek V4 arrives with API surge pricing. Practical takeaways for Australian businesses throughout. Visit www.kinsoft.com.au to talk through your security and IT needs. Sources: CNBC (Alphabet/Tesla/Intel earnings, AMD-Anthropic), AMD investor relations, SecurityWeek & Rapid7 & Qualys (wp2shell), The Hacker News & BleepingComputer (FakeGit/AgentBaiting, Anubis-Fairlife), DeepSeek API reporting.

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.

So last week, a single piece of software automatically generated a highly sophisticated strain of malware. Yeah. And then it actually recommended that malware to a human developer and honestly nearly infected an entire enterprise network. It's wild. We have officially entered the era of the weaponized AI hallucination. We really have. So welcome to another Deep Dive. Today we are looking at a July 2026 industry roundup that... frankly, completely shatters how we think about the current technology landscape. Absolutely. And our mission

for you today is to unpack a very uncomfortable reality, which is that artificial intelligence has somehow become the single most massive, heavily scrutinized line item. in corporate budgets. Right. And at the exact same time, it's the most volatile, rapidly evolving vector in cybersecurity. Exactly. It's a lot to process. It is a it's a fascinating and chaotic moment. And we are here today to really separate the financial hype from the operational reality, because whether you're a software developer trying to ship code or, you know, a CFO managing an increasingly

terrifying budget or just. Someone fascinated by where all this tech is heading. Right. Exactly. This is not just abstract news. We are going to dig into the hidden infrastructure costs and those invisible automated threats that are powering the tools you are using every single day. OK, let's unpack this. You have to start with the money, I think, because if you look at the July 2026 Q2 earnings reports, the market's reaction to big tech was absolutely brutal. Oh, yeah. Like Alphabet and Tesla both reported. And from

what I can see, they didn't really have a revenue problem. No, they didn't. But the market punished them anyway. I mean, why did the Nasdaq have its worst session since April if these companies are technically still growing? Well, because the narrative has fundamentally shifted. You know, it's shifted from revenue growth to capital expenditure. Capex. Right. Like Alphabet actually beat the revenue expectations. They posted one hundred nineteen point eight billion dollars. Wow. Yeah. And Google Cloud alone was up 82 percent to twenty four point eight billion. And Tesla beat on top line revenue, too. But Wall Street

isn't looking at the top line. They are looking at the bill for the infrastructure. The capex bill. Exactly. Alphabet raised its yearly capex guidance to over $195 billion. I mean, they spent nearly $45 billion in just Q2 alone, which is more than double what they spent a year earlier. And Tesla saw their capex jump 142%, which actually pushed their free cash flow into the negative. So it's not that they aren't making money. It's that they are spending it at a rate that is just

terrifying investors. Basically. Yeah. I've heard this described like buying an impossibly expensive sports car on a high interest credit card. But I actually think it's more like a city deciding to build a trillion dollar high speed rail network before anyone has even invented a train ticket. That's a great way to put it. Right. Because the infrastructure costs are astronomical, but the passenger revenue is entirely theoretical at this point. That is a much more accurate way to look at it. You're building foundational infrastructure with massive upfront costs. And for the last

two years, just telling Wall Street you were buying mountains of GPUs for AI was essentially a major phrase. You just got a stock bump for participating. Exactly. But these July data points show the grace period is over. Institutional investors, CFOs, boards of directors, they are demanding actual return on investment. They're looking at that high -speed rail network you just built and asking, where are the paying passenger? I do want to push back on the doom and gloom slightly there. Is this actually the bursting

of an AI bubble or is it just the inevitable hangover after a two -year party? Because tech has these massive boom and bust cycles literally all the time. I'd say it is a structural realignment, not necessarily a bursting bubble. It's the maturation of the market. And this matters profoundly to you, the listener, if you're trying to justify any kind of tech spending right now. Because the free money era is done. Right. The era of growth at all costs is dead. You can no longer just say AI and get a blank check from leadership. You have to prove exactly how a massive upfront

infrastructure cost is going to translate into sustainable, profitable revenue. Which naturally means that because these these big tech companies, the ones building the rail networks, are facing immense financial pressure, they're going to squeeze their suppliers. Yep. If the budget gets tight, the companies selling the physical hardware have to get creative. And that really brings us to the chip makers. Because I saw Intel had a surprisingly good quarter. But the real story seems to be the sheer scale of what AMD is trying to pull off with Anthropic. So let's touch on

Intel first, just because it sets the baseline. Intel posted $16 .1 billion in revenue, up 25%. Nice. But the crucial detail is how they're doing it. Their next generation 14A manufacturing process is ahead of schedule. And when we talk about 14A, we're talking about the 1 .4 nanometer class of microchips. Explain why shrinking that number down to 1 .4 nanometers actually matters, like to a CFO or a developer. Well, it really just comes down to basic physics. It's about packing more transistors into the exact same physical

footprint. The smaller the node, the less distance electrons have to travel. That drastically cuts down power consumption and heat generation. And heat is the enemy, right? Oh, absolutely. In the world of AI data centers, heat and electricity are your biggest operational bottlenecks. So whoever can make the most power -efficient chip wins the long game. And speaking of electricity bottlenecks, let's talk about this AMD and Anthropic deal. Because the numbers here do not even sound like they belong in the tech sector. They sound like they belong in the Department of Energy. They really do. Anthropic is planning to deploy

up to 2 gigawatts of AMD Instinct MI450 series GPUs. Two gigawatts. Right. And to put two gigawatts in perspective for you, that is roughly the electrical output of two full -sized nuclear power plants running at maximum capacity. Oh, wow. We are no longer measuring computing power in server racks. We are measuring it in grid strain, municipal water rights for cooling, and raw industrial output. And Anthropic is targeting that first gigawatt by the first half of 2027. But the hardware itself isn't even what makes this deal controversial,

is it? It's the financial mechanics behind how Anthropic... Can you break down this $5 billion equity investment AMD is making? Yeah, so this is a classic circular vendor financing arrangement. AMD is committing up to $5 billion in equity investment into Anthropic. And Anthropic, in turn, is committing to tens of billions of dollars in chip purchases from AMD. Wait, so this is basically like giving a friend a $50 gift card to your own restaurant just so they'll come in and order a $500 meal. Exactly. Instead of organic

market demand driving sales, AMD is essentially using its own capital to subsidize its biggest. customer. Why does that make regulators so nervous though? If I invest in a startup and they use my money to buy my software, isn't that just a tight partnership? Because it creates a mirage on the balance sheet. Regulators and analysts watch these deals closely because it artificially inflates demand metrics. Oh, I see. It makes it very difficult to tell how much real world demand there actually is for these AMD GPUs. If AMD is padding its Q3 revenue numbers with

money it effectively gave to Anthropic and Q1, well, it obscures the actual health of the market. So they're just moving money around to look good. Essentially, it's a massive strategic play to break Nvidia's monopoly, but it shifts an immense amount of systemic risk onto the books. So we have this perfect storm brewing. Companies are spending billions on hardware, but investors are demanding immediate ROI. That creates immense pressure on developers to deploy software faster than ever to justify the costs. And when you

force developers to prioritize speed over... everything else. They cut corners on security. Exactly. Which brings us to the July 17th WordPress vulnerabilities. Right. The transition from financial pressure to acute security risk is happening at light speed right now. And the WordPress exploit is a perfect example of what happens when that pressure cooker just explodes. Let's lay out what happened. WordPress patched two vulnerabilities in its core installations. So version 6 .9 and 7 .0. And this exploit chain is nicknamed WP2

shell. Yeah, WP2. I know it involves a SQL injection flaw combined with a critical code execution bug, which allows unauthenticated remote code execution. Let's translate that for anyone not deep in the cybersecurity weeds. How exactly does this work? Sure. So a SQL injection is essentially a Jedi mind trick for a database. I love that analogy. A hacker inputs a highly specific string of text into something as simple as a regular website search bar, and it tricks the underlying database into executing malicious commands instead

of a normal search. And once they bypass the database, this specific exploit allowed them to drop a web shell. And a web shell is? Think of a web shell as a hidden backdoor interface. It lets the attacker control your server directly from their own web browser, from literally anywhere in the world without needing a password. Yikes. Yeah. That is what unauthenticated remote code execution means. It is complete absolute takeover. And WordPress forced automatic updates because they knew how severe this was. But here's the part that I find so fascinating and terrifying.

Within 48 hours of that patch being released, it was being mass exploited in the wild. Yep. Like if your organization runs WordPress, you need to verify you're on version 6 .9 .5 or 7 .0 .2 right now. But I do have a question here. Sure. We keep hearing the AI is this ultimate cyber bogeyman. But a 48 -hour exploit window, isn't that just sloppy coding on WordPress's part? I mean, software has always had bugs. I actually disagree that this is about sloppy coding. To be fair, bugs happen in complex systems. You

know, they always have. The fundamental issue here is the asymmetry of time. And yes, AI is entirely responsible for this. Really? How so? Walk me through the mechanics of that 48 hour window. Historically, when a company like WordPress released a security patch, human hackers had to download it. They had to manually read the new code, compare it line by line against the old code. That takes forever. Exactly. They had to deduce exactly what vulnerability the developers were trying to fix and then manually write a custom script to exploit that missing validation.

That process took days, weeks, sometimes months. So they were basically reverse engineering the bandage to figure out. where the wound was that's a great way to put it but today ai assisted vulnerability discovery completely collapses that timeline an attacker simply feeds the old code base and the new patch into a large language model and the ai just does the work for them instantly it diffs the code identifies the logical flaw being patched and generates a working python script to exploit it What used to take three weeks of intense human labor now takes three

seconds of compute. That is terrifying. It is. The actionable takeaway for you here is that the old enterprise IT standard of patching on Monday is dead. Security is now a race measured in hours. That makes a lot of sense. The machine speeds up the lock picking process. But what happens when the machine itself is the lock being picked? Because hackers aren't just using AI to break traditional software anymore. They are directly poisoning the well of the AI tools we rely on. Oh, absolutely. Let's look at this new threat vector researchers are calling agent baiting.

This is arguably the most insidious development we've seen this quarter because it exploits trust rather than just code. Right. So the researchers found roughly 7 ,600 malicious GitHub repositories created by over 6 ,000 fake profiles. All of them were pushing InfoStealer malware. But over 800 of these repos were specifically impersonating AI agent skills and MCP servers. Now, for the listener. An MCP, or Model Context Protocol Server, is essentially the bridge that allows an AI coding

assistant to interact with external data, like hooking into your corporate Gmail, WhatsApp, or Docker containers. Right. And businesses are racing to adopt agentic AI that doesn't just write text, but actually takes actions in your environment. These MCPs are the hands and feet of the AI. So the hackers built fake malicious integrations for these tools. And the AI coding assistants that developers use actually discovered these fake repositories and recommended them to the human developers. It feels like we're hiring a brilliant assistant who is just wildly

gullible. If we can't trust our AI assistants to vet the code they recommend, doesn't that entirely defeat the productivity pitch of agentic AI? Well, I have to push back on the idea that the AI is gullible, actually. Why is that? I think that implies the AI has some baseline capacity for human skepticism. which it just doesn't. The AI is doing exactly what it was programmed to do, evaluate syntax and structure. The problem isn't the AI, it's our flawed expectation that AI possesses intuition. Okay, so how did the attackers trick it then? They engineered the

malware for the algorithm. An AI coding assistant evaluates a GitHub repository by looking for syntactical correctness, well -formatted readme files, comprehensive documentation, and standardized file structures. So they just made it look really professional. Exactly. The attackers know this, so they wrote their malware, but they wrapped it in flawless corporate -grade presentation. The AI looked at the pristine documentation, recognized all the correct formatting markers, and determined it was high -quality code. Unbelievable.

It then flagged it to the human developer saying, hey, this integration looks perfect for your project. You really cannot outsource your zero -trust architecture to a language model. You absolutely cannot. The supply chain for these AI integrations is now a primary attack surface. You have to vet a repository recommended by an AI with the exact same rigor, if not more, than you would a human third -party vendor. And the consequences of failing to vet those integrations are just... catastrophic because when developers blindly trust an ai to pull in external repositories

they are effectively handing the keys to the kingdom to whoever wrote that code right which brings us to the very real Physical consequences of these vulnerabilities. Squeezed budgets and rushed software deployments are leaving massive blind spots. And old school extortion gangs like Anubis are exploiting them. It really creates a chaotic operational reality where businesses are fighting a war on two entirely different fronts. Because on one side you have brutal extortion. Right. The Anubis ransomware gang recently claimed they successfully attacked Coca -Cola's Fairlife

dairy business. They claimed they encrypted the infrastructure, stole a terabyte of data, and set a July 27th leak deadline. Now, Coca -Cola declined to comment beyond their initial disclosure, so we only have the criminal's claims. But Anubis is unique, right? They don't just lock your files. No, they don't. They are notoriously dangerous because they frequently deploy a data wiper. A data wiper? Yes. A traditional ransomware gang encrypts your data and sells you the decryption key. A data wiper destroys the data at the sector level, making it completely unrecoverable regardless

of whether you have a key or not. Wow. How does that change the dynamic for a company that gets hit? It completely alters the negotiation leverage. If a data wiper is involved, paying the ransom doesn't guarantee you get your operational capacity back because the data might simply not exist anymore. It shifts the threat from data theft to total operational destruction. That is brutal. So you're fighting off literal data wipers on one front, but on the other front, you're... Internal software might just break because your AI vendor flipped a switch and decided to change

their business model overnight. Yes. Let's talk about what happened with DeepSeq last week. The DeepSeq rollout is a masterclass in hidden operational risk. They made their V4 family of large language models generally available. Which sounds great for developers, except they immediately retired their legacy API endpoints on the exact same day. No transition period at all. Zero days notice. In traditional enterprise software, if a vendor is deprecating an API, they give you six months, maybe a year to migrate your system. Right. That's

standard practice. But in the frontier AI market, DeepSeek just flipped a switch and it instantly broke live requests for developers who hadn't migrated that very morning. And it actually gets worse than just forced migraines. Secondary reports show that DeepSeek is actively utilizing Uber -style surge pricing in the LLM API market. Yeah. Their compute rates were roughly doubling. during Chinese business hours. Why is that happening and how does it impact the developers relying on it? It happens because going back to our earlier point about gigawatts and infrastructure, AI

compute is physically constrained by the electrical grid and hardware availability. Right, the data centers. During peak hours, demand outstrips supply. So DeepSeek uses dynamic pricing to throttle usage and prioritize the highest paying tasks. Wow. But for the developer, this is devastating. If your entire business model relies on a low cost API and that vendor introduces surge pricing that suddenly doubles your server costs during your peak operating hours, your profit margins just evaporated instantly. Cheap AI APIs do not

mean flat rate forever. So what does this all mean? Let's bring this all together for you. The thread running through everything we've discussed today, from Wall Street panicking and demanding ROI on Alphabet's massive spending, to Anthropic securing nuclear levels of power with AMD's money, to WordPress being exploited in 48 hours, and AI APIs breaking businesses with surge pricing, is that artificial intelligence has fundamentally matured. It has. But it hasn't matured into a stable utility like electricity or water. It has matured into both the biggest, most scrutinized

line item on your budget and the absolute fastest moving, most unpredictable part of your threat landscape. It is the foundation we are building on, but that foundation is actively shifting beneath our feet. Yeah. And before we go, there is a final lingering question I really want you to mull over. It takes all these threads, the automated exploits, the gullible agents, the surge pricing, and weaves them into our near future. Okay, let's hear it. We know our AI coding assistants can be tricked by polished documentation into recommending malware today. But what happens

tomorrow when autonomous AI agents start negotiating their own software contracts on our behalf? Oh, man. Or what happens when those agents are authorized to automatically pay API surge pricing bills without human oversight? Could a hacker literally bankrupt a company not by stealing its data, but by tricking its AI into an infinite loop of surge? priced API calls. Wow. Death by a trillion automated microtransactions. It just goes to show you can't build a trillion dollar high speed rail network without first figuring out who is

actually allowed to drive the trains. Thank you so much for joining us on this deep dive. We warmly invite you to keep questioning the tech you rely on every day. We'll see you next time.

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