My company has taken part in previous infra/STRUCTURE summits, but this was the first time I’ve attended, and honestly, I wasn’t sure what to expect. You know how it goes: sometimes industry events live up to the hype, and sometimes they do not. For me, infra/STRUCTURE was definitely worthwhile.
The summit is organized by Structure Research, and focuses on cloud, datacentres and edge – the infrastructure that makes the internet tick. The theme this year was ‘recalibration’ and the need for the infrastructure industry to focus on collaboration and partnership, rather than taking a ‘supplier and vendor’ approach.
I attended a few sessions, met some interesting people, and took part in a panel session about the future of cloud, which took an unexpected left turn. More on that in a moment, but first, here are some other key takeaways.
Is anyone NOT talking about AI?
The agenda was so packed it was impossible to take in every session, but if there was one common theme, it was AI. It felt like every single session was dominated by the AI topic, and almost every booth, too.
The need for AI and GPU workloads is driving demand for non-hyperscale cloud services and smaller datacentres again. These smaller, more agile infrastructure providers can offer infrastructure that is more localized and specialized for edge computing, blockchain, AI (machine learning/ language models), other GPU-specific workloads, and even application workloads that need high performance bare-metal.
The expectation was that smaller datacentre demand would decrease, but there seems to be a cyclical trend where new innovation requires these smaller datacentres. These workloads are driving growth in the infrastructure market: it’s net new business, not just repatriation of existing workloads. This bucks the industry expert predictions that consolidation is killing this market for smaller infrastructure providers. However, consolidation was itself a hot topic, which I’ll get to shortly.
Datacentre power problems
There are two problems for smaller infrastructure providers, and they are both about power.
Emerging and experimental AI-based applications and services, which depend on large amounts of GPU/APU hardware, require more power per square foot in the datacentre than most existing facilities were built for. A retrofit movement is taking place, but this is often constrained by the power utility company.
New datacentre builds are still happening, of course, but most of those new facilities are being built to service just one company. I talked with an investment firm rep at infra/STRUCTURE who said, the reality now is that they are backing entire datacentre build projects on a single vendor contract – a large hyperscale public cloud provider, or a large SaaS company such as Facebook, for example.
This goes against the traditional model of building a carrier- or customer-neutral facility with the expectation that multi-tenancy will justify the construction cost. The purpose for which physical internet infrastructure is being built has changed dramatically. It’s not just an electrical power problem, it’s an empowerment problem with access to investment. Here’s where we get to the consolidation challenge.
Next-generation cloud? Left turn, Clyde
I was fortunate to be invited to join a panel session, along with Rocket.net’s Ben Gabler, Vultr’s DJ Johnson, and Carrie Fisher from Liquid Web. Moderated by Structure Research’s Philbert Shih, it was the last session of the day in the cloud and MSP track – a real graveyard slot after such a packed agenda on day one of the summit, but it turned out to be quite the popular discussion.
The topic was ‘Next-Generation Mass-Market Cloud Services” and for the first 15 minutes we talked about the fact that MSPs are less interested in the underlying infrastructure and more interested in the services they can offer their customers, which requires that datacentres modernize and focus on providing simple ways for service providers to sell solutions to end users.
No great surprises there, but during the discussion it suddenly hit me: all this talk about the exciting AI future for the infrastructure industry, and challenges for MSPs and smaller infrastructure providers, that’s all well and good – but it doesn’t address the real issue with internet infrastructure today. Which is, that consolidation is killing the industry.
Gartner’s recent estimate is that 83 per cent of the cloud infrastructure market is owned by three companies. It’s bad for the industry, it’s bad for the end user, it’s bad for business. Here’s where the panel session took a left turn, because that’s all we talked about from then on.
The problem is choice
There are many reasons why the world needs independent infrastructure providers, and independent cloud providers. As discussed at the summit, emergent technologies like AI often require distributed, highly localized infrastructure. From a legal and compliance standpoint, organizations need to build local infrastructure and local providers into their IT strategy.
From a commercial and end user perspective, competition is healthy, not just for innovation but for optimizing price, service and support – and, at the most fundamental level, people need choice, not monopolies. It was a discussion that quickly veered into broader topics, such as net neutrality and the end goal of a capitalistic model for the cloud industry.
Did the audience object to this change of direction? The panel was scheduled to last 30 minutes. We ended up speaking for twice as long, with plenty of audience participation during the session and afterwards over drinks.
When it comes to the future of cloud, and the future of internet infrastructure as a whole, the effect of consolidation is the real topic the industry needs to address. It’s a fundamental issue. It’s an existential issue.
There are companies and organizations working to solve this problem – the company I work for, Virtuozzo, is one. What does the solution look like? Neatly this brings us back to the overall theme of infra/STRUCTURE 2023, that the industry needs to think in terms of partnership and collaboration, instead of having a supplier/vendor mindset.
Smaller infrastructure providers are perfectly placed to enable AI and other new technologies, to offer choice of price and service and support, and local locations for compliance. However, a small provider is still a small provider: resources are finite, locations are limited, and they are always going to come up against a “who the hell are you?” mentality when customers compare them to bigger, better-known infrastructure brands.
What’s needed is a simple way for businesses to access all of the smaller providers at once – so that collectively, those smaller providers can offer not just the same level of scale you can get from the large public clouds but do so while also maintaining the things that make them unique. Like the opposite of the Borg: you become part of a larger collective, but your technological distinctiveness complements that collective, rather than being assimilated by it. As part of a collective, it’s easier to compete, and it’s easier to get funding and support for the infrastructure needed for the next generation of AI services, too.
This is something I’ve believed for a while, and it’s something we’re already working on at Virtuozzo. But for me the key takeaway from infra/STRUCTURE was the sheer volume of the affirmation I heard from other industry professionals. The summit was often about AI, but the echoes that stayed with me were about the vital importance of protecting and enabling our vibrant, multi-dimensional service provider community, to protect the future of our internet. Let’s make it happen!