Alphabet Cloud Boom, Negative Cash Flow & TSMC Hike

Samantha: Welcome to The State of Tech, The European Edition, Thursday July twenty-third, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Alphabet's cloud boom collides with a cash flow warning, AMD unveils its next wave of AI chips, Washington and Beijing plan their first formal AI talks under Trump, TSMC signals a chip price hike into 2027, and to close out, two things you can actually use today: the encrypted messenger Signal for desktop workflows, and the free travel-photo cleanup tool Lensless. Let's start with Alphabet.

Alphabet's cloud business roars ahead, but AI spending drags free cash flow into negative territory.

Samantha: Alphabet just posted its second-quarter numbers, and it's a genuinely split picture. Google Cloud blew past revenue expectations, and the order backlog for cloud services now sits above five hundred billion dollars. That is an eye-watering signal of demand from businesses lining up for AI infrastructure. But at the same time, free cash flow turned negative, because Alphabet is spending enormous sums on data centres, chips, and everything needed to actually deliver on that backlog. Shares slid in after-hours trading on Wednesday as investors tried to make sense of the tension.

Bob: And that tension is the whole story. On paper, a half-trillion-dollar backlog looks like a golden ticket. In practice, converting it into revenue requires building physical capacity at a pace we have not really seen in tech before. Alphabet is essentially telling the market: trust us, the spending pays off later. Investors are asking how much later, and what happens if AI demand cools or if margins on cloud AI turn out thinner than expected. The scrutiny is not just about one quarter, it is about whether the entire hyperscaler business model still works when capital expenditure runs this hot.

Samantha: For European listeners this matters on two levels. First, European companies are among the customers filling that backlog, so the pricing and availability of Google Cloud AI services here depend on how this build-out plays out. Second, European tech firms watching from the sidelines have to make their own calls on AI spending. If even Alphabet, with its scale and cash pile, is taking a hit to free cash flow, smaller European players face harder choices about whether to build, buy, or partner. The bottom line is that the AI infrastructure race is now openly reshaping balance sheets at the very top of the industry, and that ripples all the way down.

Bob: One more thing worth noting. The backlog figure implies multi-year commitments, which means Alphabet's customers are locking themselves into Google's stack for the long haul. For European regulators already wary of cloud concentration, that is a data point. It suggests the dependency question is not going away, it is deepening, quarter by quarter.

AMD unveils its next-generation AI chips and rack systems to challenge Nvidia head-on.

Samantha: AMD is hosting its Advancing AI 2026 event today, with CEO Lisa Su on stage unveiling the company's next-generation AI lineup. The headline products are the new EPYC server processors and the Instinct MI400 series accelerators, which together power a new rack-scale AI system called Helios. In plain language, AMD is not just selling chips anymore, it is selling ready-to-deploy AI computing rooms, aimed straight at cloud providers and large enterprises that currently buy almost exclusively from Nvidia.

Bob: And that competitive angle is really the point. Nvidia has dominated AI training and inference hardware for years, and its pricing power reflects that. AMD's pitch today is that Helios gives buyers a credible second option, with performance in the same league and, crucially, availability that is not stuck behind Nvidia's waiting list. If AMD can deliver at scale, the entire economics of AI infrastructure shifts. Cloud giants get leverage in negotiations, and prices for AI compute could actually stop rising quite so steeply.

Samantha: For Europe, this hits at a practical level. European AI companies and research groups have been vocal about being unable to get enough Nvidia hardware, and about the cost when they do. A viable AMD alternative means European cloud providers, from the big French and German players to specialised AI labs, could build out capacity faster and more affordably. It also matters for sovereign AI projects, where governments want compute inside European borders and are less fussed about which logo is on the chip, as long as it works.

Bob: The catch, of course, is software. Nvidia's dominance is not just about silicon, it is about the years of tooling and libraries developers rely on. AMD has been closing that gap steadily, but today's announcements will only matter if the software story convinces engineers they can switch without pain. That is what analysts and European buyers will really be watching over the next few months.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

The United States and China plan their first formal AI policy talks under President Trump.

Bob: Washington and Beijing are reportedly preparing to sit down in September for bilateral talks specifically focused on artificial intelligence policy. The agenda centres on the safe management of advanced frontier AI models, the kind of systems whose capabilities are outpacing the guardrails around them. This would be the first official AI meeting between the two countries under President Trump, and it signals that both governments are, at least publicly, worried enough about the pace of development to talk to each other about it.

Samantha: The framing here is careful. This is not a treaty negotiation, and neither side has hinted at binding commitments. What is on the table is a shared vocabulary for risk: what counts as a frontier model, what kinds of incidents should trigger notification, and whether there is any common ground on safety testing before release. Given the broader trade tensions between the two countries, even agreeing to meet is notable. It suggests AI safety is one of the few areas where cooperation still looks politically possible.

Bob: For Europe, this is a moment to watch closely rather than participate directly. Brussels has been the loudest voice globally on AI regulation, with the AI Act now in force and setting a template many other countries reference. If Washington and Beijing start converging on their own shared principles, even loosely, that could either reinforce the European approach or sideline it. European officials will want to know whether the talks build on existing multilateral efforts, or whether they carve out a US-China track that leaves Europe reacting from the outside.

Samantha: The other question is what industry does with this. Frontier AI labs, both American and Chinese, have been calling for clearer international rules for over a year. If the September talks produce even a modest framework, it gives companies something concrete to align their safety practices against. And for European AI firms, it could reduce the risk of being caught between three incompatible rulebooks.

TSMC will raise chip prices by up to ten percent in 2027, sending costs through the entire tech supply chain.

Bob: Taiwan Semiconductor Manufacturing Company, the world's largest contract chipmaker, plans to raise its prices by up to ten percent starting in 2027. Sources close to the company say the increase covers both older and cutting-edge production, and applies broadly across customers. The reasons cited are rising costs for raw materials, more expensive manufacturing equipment, and the enormous bill for building new plants outside Taiwan, particularly in the United States, Japan, and Germany. This comes right after TSMC posted a record second-quarter profit, largely driven by AI chip demand.

Samantha: The interesting part is not that TSMC is raising prices, it is who ends up paying. TSMC makes chips for almost every major name you can think of, from Apple and Nvidia to AMD and Qualcomm. A ten percent bump at the foundry level cascades through the entire electronics industry. It shows up eventually in phones, laptops, cars, and increasingly in the servers powering AI. Given that TSMC already has customers queuing for capacity, buyers do not have much room to push back. It is genuinely a seller's market for advanced chipmaking.

Bob: For Europe, the impact runs in two directions. On one hand, European carmakers, industrial firms, and consumer electronics brands rely heavily on TSMC and will face higher input costs, which may show up in product pricing over the next couple of years. On the other hand, the fact that TSMC is spending so much on overseas plants, including its Dresden facility in Germany, is exactly what European industrial policy has been pushing for. The price rise is partly the cost of that diversification. Europe wanted chips made closer to home, and this is what closer to home costs.

Samantha: There is also a longer game here. If TSMC can raise prices and still keep customers, it signals just how few real alternatives exist at the leading edge. Samsung and Intel have been trying to close the gap for years, but neither offers a broad enough menu to seriously challenge TSMC yet. For European firms depending on this supply chain, the message is that concentration risk in chipmaking is still very much a live issue.

Signal's new desktop features turn the encrypted messenger into a serious work tool.

Bob: To close out, two things worth trying today. First, the encrypted messenger Signal has been quietly rolling out a set of desktop improvements that reviewers are calling a genuine upgrade for anyone who works across a laptop and a phone. Signal is best known for its privacy credentials, with end-to-end encryption by default and no phone number visible to contacts if you set up a username. What has changed is the desktop experience: much smoother file sharing, better search across old conversations, and a redesigned interface that finally feels like a proper work tool rather than a phone app awkwardly stretched to a bigger screen.

Samantha: The reason it is worth a look now, especially for European listeners, is that Signal is run by a nonprofit foundation and does not sell ads or scan messages for training data. That matters if you have been looking for a middle ground between WhatsApp, which sits inside Meta, and Telegram, which has its own privacy questions. Users report Signal now handles larger files without breaking a sweat, and the group chat features have caught up enough that small teams can use it for actual coordination, not just casual chat.

Bob: It is free on Windows, Mac, Linux, iPhone, and Android, and you can install the desktop version in a couple of minutes. The one thing to know is that you still need to link it to a phone number to get started, though you can hide that number from others afterwards. For anyone tired of their messaging app doubling as an ad platform, this is a low-effort switch that costs nothing to try.

Samantha: And the practical test many European users apply is whether family and colleagues will actually follow you there. Signal's user base has grown steadily across Europe, particularly among journalists, activists, and privacy-conscious professionals, so the network effect is less painful than it was a few years ago. Give it an afternoon, move one or two conversations across, and see how it feels.

Lensless is the free web app that cleans up your travel photos in seconds without uploading them to a big cloud.

Bob: The second thing to try today is Lensless, a free web-based photo tool that has been getting a lot of attention among travel photographers and casual holiday snappers alike. What it does is simple but genuinely useful: it removes tourists, background clutter, and small distractions from your photos, and it does most of the processing directly in your browser rather than uploading everything to a distant server. So that photo of the Trevi Fountain with fifty strangers in the frame can become a much cleaner shot in under a minute.

Samantha: The technology behind it is not new, background removal and object erasing have been around for years in expensive editing suites, but Lensless has stripped it down to something anyone can use without a subscription or an account. Reviewers highlight that the results are surprisingly clean for a free tool, particularly on landscape and architecture shots where the software has clear edges to work with. It struggles more on complex crowd scenes, which is fair to know before you expect miracles.

Bob: For European users heading into peak holiday season, it is a handy addition to the phone. You can open Lensless directly in your mobile browser, drop in a photo, and export the cleaned-up version without ever installing an app. And because much of the processing stays local, you are not handing your holiday album to another platform for training data. Free, no login, works on any device with a modern browser. Worth ten minutes this evening to see if it saves your favourite shots.

Samantha: The one honest caveat is that free tools tend to add limits over time, so today's generous version may become tomorrow's paywall. But right now, there is nothing to lose by giving it a spin.

Samantha: Today we covered: Alphabet's cloud boom shadowed by cash flow concerns, AMD's next-generation AI chips going after Nvidia, US-China talks on AI policy, TSMC raising chip prices into 2027, and two things to try yourself: the encrypted messenger Signal on desktop, and the free photo cleanup tool Lensless.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.