The Neoscaler Procurement Gap: Why the Network Shouldn’t Be the Last Thing You Buy
KEY TAKEAWAYS
- Many neoscalers buy GPUs, land, construction, and power first, so the network is bought last and often under real-time pressure.
- Small networking teams often design the network and run the procurement at the same time, so specifications test whether capacity can be delivered but not cost, contract structure, or long-term exposure.
- Contract terms do not match: GPU compute is increasingly sold on six-month terms, while fibre and lit services typically run three to five years and dark fibre/IRU commitments run 10–30 years.
- Short terms cost more, and the same capacity can cost nearly double on a six-month term compared with a three-to-five-year term.
- The fix is commercial discipline applied before the deal is signed, not more engineering talent, and European neoscalers may be more exposed because there is less evidence of specialist buying teams there.
3 MIN READ
As my interest in the impact of AI on procurement has grown, I've come across many articles about the growth of neoscalers or neocloud operators (from now on, I'll refer to them simply as neoscalers).
Looking at how some of these neoscalers seem to buy their infrastructure, a pattern quickly emerges. GPUs first, then land, construction, and power, and finally the network. Because the network is last on the priority list, it is often procured under real-time pressure by the time anyone gets to it.
This isn't a criticism of the people responsible for purchasing. Most neocloud networking teams consist of a small number of technically skilled members who now have to juggle two tasks at once: designing the architecture (including network capacity) and carrying out procurement, since no one else is available to take on those tasks. For a company operating at the pace neoscalers do, this is a sensible approach to staffing the first few months of running the business. It is also at this point that the gaps begin to appear.
In my consultancy work, I am starting to see many instances where the specification itself reveals those gaps. One instance I came across recently required a neoscaler’s potential network supplier to provide capacity across three distinct paths by a named delivery date, with no term specified when the request went to market. That's not an unusual ask in this market. It's often a sign the request was written by someone testing delivery certainty, because that's what an engineer tests for, rather than by someone weighing cost, contract structure, or long-term exposure, because that's what a buyer weighs. Both questions are important, but the request to the capacity market only asked one.
The same gap reappears with contract-term requirements versus cost expectations, and here the exposure is more financial than technical. GPU compute is increasingly sold on six-month terms, so the neoscaler thinks in these timeframes. The capacity contracts underpinning a neoscaler’s infrastructure typically run for at least one year, and often up to five years or more. Fibre leases and lit services typically run for three to five years at a reasonable price. Dark fibre and IRU commitments can run 10 to 30 years. It gets more complex where suppliers are buying long-term capital dark fibre/IRU commitments from the underlying cable owners to get access to sites where neoscalers reside and then trying to make money selling to a customer base with a very short-term view of capacity contracts.
There are cases where the same capacity, quoted on a three-to-five-year term, is only available on a six-month term at nearly double the price.
This issue does not need more engineering talent to fix. It needs a different discipline applied at the point the deal is done and before it gets signed. These are procurement questions, not technical ones, and they rarely get posed when the person asking is the same person who is trained to design the network.
The pattern is not wholly universal, though. Some North American neoscalers show a strong commercial/procurement capability through having (or buying in) a specialised buying team. The equivalent evidence from Europe is limited. While the publicly available sample size of sales announcements/public RfPs is small, it suggests European neoscalers may be more exposed to these risks.
None of this is new. It's the same story I’ve watched play out, over the last 30 years, across plenty of fast-growing businesses in the wider telecoms world. The business is in startup mode, and the procurement function doesn't get built until volumes are large enough to force the need; by then, the damage from early deals is already done, or the advantage of getting it right from day one is already lost. It is also the case that many procurements for capacity are made as a one-off activity at the start of the build, with a medium-to-long-term timeframe, and so no one wants to hire a full-time in-house procurement team just to make the buy and then sit around until a new piece of capacity is needed or a term renewal needs to be considered.
External procurement specialists are eventually engaged. The practical issue is how much is spent, or how much value is conceded, before that engagement occurs.
Procurement Health Check
This article outlines the use case that our Procurement Health Check is designed specifically to address. It involves a comprehensive review of various elements, including: organisational structure (including how procurement interfaces with the rest of your business), strategy, procurement processes, technology, and supplier relationships. From this analysis, we identify areas for improvement and ensure value for money. We can also conduct a benchmarking review of your connectivity costs to reveal areas for cost reduction.
Additionally, your organisation can engage Cambridge Management Consulting’s market knowledge and expertise using our Procurement as a Service capability to support your teams until your internal capabilities are justified and in place.
Find out more on our
service page for Procurement & Commercial.
About the Author
About Us
Cambridge Management Consulting (Cambridge MC) is an international consulting firm that helps companies of all sizes have a better impact on the world. Founded in Cambridge, UK, initially to help the start-up community, Cambridge MC has grown to over 200 consultants working on projects in 25 countries. Our capabilities focus on supporting the private and public sector with their people, process and digital technology challenges.
What makes Cambridge Management Consulting unique is that it doesn’t employ consultants – only senior executives with real industry or government experience and the skills to advise their clients from a place of true credibility. Our team strives to have a highly positive impact on all the organisations they serve. We are confident there is no business or enterprise that we cannot help transform for the better.
Cambridge Management Consulting has offices or legal entities in Cambridge, London, New York, Paris, Dubai, Singapore and Helsinki, with further expansion planned in future.
Contact Form
Subscribe to our Newsletter
Blog Subscribe
SHARE CONTENT
