Cloud Wars Live with Bob Evans

In today’s Cloud Wars AI Minute, I break down Gartner’s $2.59 trillion AI infrastructure spending figure and why rising compute and memory costs are shifting where the economics of the AI boom are landing.
Highlights
00:11 — Let's talk about this new Gartner number that just came out. Gartner is claiming that it's $2.59 trillion in CapEx expenditures that we're starting to see companies investing in the AI infrastructure space. Well, this is a very big number, and in an earlier video, I covered the fact that now we're starting to see cash flow not being able to keep up.
01:02 — If we take a look at Nvidia for a second, that is going to be running a lot of this compute on its hardware, you're going to see that they're at $89 billion in one quarter. Now, that's their revenue in just one quarter alone, and that's up 117 percent year over year. This means that they're demanding prices and they're manufacturing this in a market that the demand is far exceeding their ability to actually produce equipment.
01:36 — And you're also seeing 400% DRAM prices going up since the start of 2024, and so when you combine these two different pieces together, what we're seeing is that the revenue is really being made by the hardware manufacturers and the ability to do the compute. This is where all revenue seems to be going right now.
02:15 — Gartner has even predicted that in 2026 we're going to see the trough of disillusionment happen in the AI space. Now, what this really means is that we're going to see the hype cycle bring it down a bit to be able to go into what is the actual reality of AI return on investment, and so we should start seeing this hype cycle come down.
03:22 — And so this is going to be a very interesting thing for us to continue to watch throughout 2026, and it should open up some opportunity and hopefully maybe help us reduce the demand that we have on both compute and memory infrastructure, allowing this to normalize a bit.

Visit Cloud Wars for more.

In today’s Cloud Wars Agent and Copilot Minute, I break down Microsoft’s rollout of OpenAI’s GPT-6 Astra in Foundry and why enterprise controls could be a major differentiator for agentic AI.
Highlights
00:09 — So a quick recap before I get into this story: at the start of September, OpenAI launched GPT-6 Astra, which it described as the world's most intelligent and aligned model. Now, I'd urge you to check out the promotional video that accompanied that launch, because it's pretty mind-blowing in terms of what this model can do.

00:31 — It's been a shift away from Q&A or chat and into actions. So, we're talking computer use, software engineering, browsing, and so on. Now, the reason we're talking about this today is that shortly after launch, Microsoft announced that it had made this frontier model generally available for all customers in Microsoft Foundry.
00:53 — Now, in practice, this means that customers can use the model to complete multi-step, agentic work with open-ended goals. Microsoft is really pushing the idea of computer use and how transformative this can be for its enterprise clients.
01:09 — And the model is pricey, with the standard global tier starting at $10 per million input tokens and $50 per million output tokens, short context. Long context is $20 per million input tokens and $75 per million output tokens.
02:29 — Foundry complements OpenAI's safety alignment and other factors with security, safety, and compliance capabilities, including Microsoft Entra identity and access management, encryption in transit and at rest, private networking options, role-based access controls, content filtering, safety evaluations, monitoring, and, of course, governance tools. And that is what I think will draw people to Microsoft in this case.

Visit Cloud Wars for more.

In today’s Cloud Wars AI Minute, I break down the explosive growth in AI infrastructure spending and the growing pressure on hyperscalers as CapEx begins to exceed operating cash flow.
Highlights
0:14 — The thing I'm going to talk about today is the fact that we are now seeing over $700 billion in AI buildout as far as the investment in CapEx that's starting to be spent. Well, this is really an interesting situation because we're starting to see that the actual CapEx expenditures are exceeding the cash flow funding it at this point.
0:57 — Amazon is saying alone in 2026 it had to raise to $220 billion in CapEx expenditures, due to memory cost as the cited reason. So, now the question becomes: Will the memory shortage start to push back a little bit in this because of the fact of where these numbers are going?
2:09 — How do we continue to fund these things where we don't actually see the cash flow being able to be a return? Now, currently, right now, we're sitting at $2.59 trillion total worldwide AI spend, and this is up 47% from 2026.
2:53 — We're either going to see them have to borrow money, or they're going to have to go into a situation where they tune it back a bit to be able to deal with the fact that it's starting to exceed their cash flow. It'll be an interesting space, and it will be something that we'll continue to watch as we continue to go forward.

Visit Cloud Wars for more.

In this Cloud Wars Minute, I dive deeper into comments from Oracle's Q1 earnings call on its philosophy, position, and strategy where agents meet applications.
Highlights
00:16 — I think Oracle has, by far, been speaking with the most clarity about what the interplay between apps and agents will be. This is incredibly important for customers. Just a couple of years ago, Microsoft CEO Satya Nadella turned the whole apps world upside down when he suggested that the arrival of AI, agents, and copilots would hollow out applications.
00:46 — During the Q1 earnings call, CEO Mike Sicilia made remarks about Oracle's view on applications and where apps and agents collide. He focused on the notion that Oracle's suites of applications are moving into industry suites that seamlessly blend agents and apps together.
01:48 — Sicilia further stated that AI is an accelerator for packaged apps, not a replacement. He cited a number of ways that acceleration is happening. With this blend of apps and agents, employees will shift to overseeing agents, resolving exceptions, and using human judgement to solve complex decisions that arise. This very tight blending of what they're calling "fusion" agentic applications will make that possible.
02:21 — It's important to remember Nadella's comments on a podcast interview two years ago as he described this, saying that AI is going to take command and applications will be rendered much less valuable. He said they would be going after this with agents and copilots aggressively, with hopes of collapsing it all, referring to the current state of the role of applications at the time.
03:55 — Oracle has done the best job of consistently describing where apps end and where agents begin as well as where agents complement what apps are doing.

Visit Cloud Wars for more.

In today’s Cloud Wars AI Minute, I break down the rise of AI model fatigue and why the industry’s rapid release cycle may be moving faster than buyers can keep up.
Highlights
00:16 — The thing we’re going to talk about today is going to be model fatigue. And this is a real thing that’s starting to happen in the industry as these models keep coming out, and they come out so quick that you’re starting to see that the race of the releases are just outrunning the buyer’s ability to be able to consume them.
00:37 — And just to give you some perspective of some key concepts that kind of showcase this is five of the key frontier models all released new versions within four days of one another, and when we start looking at this and having this model fatigue, you’re actually starting to see the labs that build these different frontier models starting to ask for the slowdown of the development of AI.
01:13 — What we really are starting to get into is going to be that we need to make sure that we have the ability for people to consume these things and understand the value. You’re also starting to see that there is also a need for deeper integration and depth, and data quality and domain fit are actually starting to be much more important than just releasing a new version of the model.
01:48 — The other perspective of this is not only is it causing a problem in being able to adopt, it’s also causing a problem in ability to actually service the needs. So we saw that Astra, which is GPT-6 Astra, literally in seven days turned off its Pro offering. And why did they do that? A lot of it just comes down to the race to produce these new models.
02:20 — And so now, what we’re seeing is that even if you make these new models, getting the capacity rolled out to be able to service the model’s demand is just not going to be able to be kept up with. So I anticipate what we will see is a slowdown in the number of models, so that it will be more consumable for the market in general.

Visit Cloud Wars for more.