Should early-stage startups pursue a single-cloud, multicloud or on-prem strategy when just starting out?
Well, the simple answer to that question is just one cloud, but in the wake of the Silicon Valley Bank collapse, redundancy has become sexy again: Who wants to be dependent on a single provider for any mission-critical activity?
TechCrunch+ is currently surveying startup founders and CTOs regarding their cloud usage and best practices.
If you’re interested in participating in this survey, we’d love to hear from you!
But it appears the main consideration isn’t redundancy — it’s actually what sort of compute load a startup has to deal with, according to a TechCrunch+ survey of several startup founders and CTOs. Notably, the founders we heard from were generally bullish on single-cloud usage for young startups, with significant caveats: If a young tech company is simply hosting software, then, to start, a single cloud will suffice, but if the company is working on AI-related tasks like training models, it may need more.
Satyen Sangani, co-founder and CEO of Alation, described when it might make sense to use something other than the cloud:
If your company needs a huge amount of infrastructure right at the beginning (say, you’re training the next large language model), it might make sense to buy hardware instead. Generally, the early days of companies are filled with experimentation, and the flexibility that clouds provide is a massive benefit in those days.
A good question at this juncture is what fraction of “AI-first” startups are training their own models instead of remixing or retooling the UI-layer of existing LLMs, for example. We reckon it’s not too high.
Regardless, after parsing answers to our first question, the next time we ask something related, we’ll amend our prompt to: When should a startup move to a multicloud setup?
For now, read on for answers to our question: Should early-stage startups pursue a multicloud or on-prem focus when they’re just starting out?
We spoke with:
- Tobi Knaup, founder CEO, D2iQ
- Mang-Git Ng, founder and CEO, Anvil
- Joe Mainwaring, director of Infrastructure, WorkTango
- Vikas Bhatia, co-founder, CEO and chief risk officer, JustProtect
- Satyen Sangani, co-founder and CEO, Alation
- Steve Mullaney, president and CEO, Aviatrix
- Ed Thompson, CTO, Matillion
- Adrian Estala, VP, field CDO, Starburst
- Shane Buckley, president and CEO, Gigamon
Tobi Knaup, founder and CEO, D2iQ: As always, it depends. But most software startups should start on one cloud and be careful not to create too much lock-in by using proprietary services so you can optimize and migrate more easily later.
If your company needs a huge amount of infrastructure right at the beginning (say, you’re training the next large language model), it might make sense to buy hardware instead. Generally, the early days of companies are filled with experimentation, and the flexibility that clouds provide is a massive benefit in those days.
Mang-Git Ng, founder and CEO, Anvil: Single cloud.
Joe Mainwaring, director of Infrastructure, WorkTango: Single cloud.
Vikas Bhatia, co-founder, CEO and chief risk officer, JustProtect: When starting out, only a single-cloud solution makes sense. That said, if at all possible, I would recommend leveraging low-code technology to validate the concept.
Satyen Sangani, co-founder and CEO, Alation: Single cloud. Early-stage startups need to reduce complexity and optimize for speed. Unless the business has more than 40% of its costs going to the cloud, it’s unlikely that using multiple clouds would provide a material enough benefit in either costs or risk reduction. If cloud costs are a relatively significant part of spend, like at companies that are now developing large-scale, generalized large language models (LLMs), then it’s possible that the answer shifts.
Steve Mullaney, president and CEO, Aviatrix: Cloud is often considered a utility, like electricity, by enterprise application teams and other consumers. However, infrastructure operations teams require specialized expertise and underlying technology to deliver the agility and perceived transparency that lines of business expect of cloud-based services.
For startups, you need something that’s there working behind the scenes, not hindering the business and not something you need to dedicate many internal resources or overhead to manage. Single cloud gives you the flexibility to scale your infrastructure to meet the business’s needs.
But as any business grows today, it will eventually be multicloud, whether due to varying department or application requirements, M&A activity, shadow IT or something else. Regardless of how they begin, startups should expect to eventually be multicloud.
Ed Thompson, CTO, Matillion: It depends on the exact goals of the startup. For example, there are lots of businesses out there that exist to help with management of workload across cloud platforms. However, assuming I am just looking for a platform for my product idea in “early stage,” I would opt for a single cloud and lean in hard on the technologies that they offer that allow you to concentrate on proving the business idea at scale. It’s all too easy to get bogged down with technology and infrastructure with the misguided view that saving a few dollars here and there will put you on solid ground for the future.
Time is the enemy of the high-growth business — speed to market is the key. Test your ideas, fail fast, learn, pivot and try again. I would always steer toward serverless tools that would accelerate development and allow you to rapidly scale without worrying about the underlying infrastructure.
Now, this approach does have downsides, namely the vendor lock-in and associated cost of going heavily into something like AWS Lambda. However, a successful idea that grows fast and is adopted at scale but costs too much to run is a much better problem to have than an idea that reaches the market too late and is overhauled by more nimble competitors.
Adrian Estala, VP, field CDO, Starburst: Starting with a single cloud provides incredible advantages. But we live in a multicloud, on-prem world, so any new startup has to account for that diversity.
The ability to integrate seamlessly across this hybrid environment should be the goal for any startup looking to provide IT/data services.
Shane Buckley, president and CEO, Gigamon: Single cloud is the easy answer, but that’s not the reality. The goal of a seed-funded company is to quickly validate, or invalidate, their business concept. Once that stage is reached, the focus shifts to building customer trust with applications that are both secure and deliver a great user experience.
While all companies begin with a single cloud, we find most companies soon land on hybrid and hybrid-cloud implementations. Capital constraints, need for agility, available cloud resources and expertise, and M&A activities all factor into how hybrid cloud infrastructures evolve.
While public cloud platforms enable rapid deployment and scalability, something that is obviously compelling for startups, the “shared responsibility model” employed by these providers means that application and workload security is the responsibility of the organization, not the provider. Just as guaranteed cloud compute performance does not automatically translate to a great application user experience, physical data center security does not guarantee end-to-end application security either. That is ultimately achieved by gaining deep observability into all hybrid-cloud communications.