Scale Extrapolation Errors in Telecoms/Tech

Recently I wrote a fairly exasperated LinkedIn post about space datacentres and why they are a great example of what I’m calling a “scale extrapolation error”. They're a common type of cognitive bias in telecoms, technology and cloud / AI domains.

Scale extrapolation assumes a technology or business model that you observe at one level can scale smoothly (linearly or exponentially), up or down, while ignoring thresholds, multi-variable optimisation, and non-linear constraints.

I've written a full Substack article on the general definitions and causes of that error class.

The space DC post pointed out that having a (potentially good) way to scale one variable or input for a complex sector does not mean all the other necessary inputs or variables also scale.

Just because solar energy is cheap in space and it’s suddenly possible to deploy hundreds or thousands of satellites does not mean that it’s a substitute for terrestrial hyperscale DCs for cloud or AI use. They also need cooling, connectivity, maintenance, interconnection, developer support, low latency, reliability, regulatory oversight and much more.

I also cited some other "X can substitute Y" scale extrapolation errors and fallacies I commonly see:

• The belief that terresrial edge computing nodes can replace hyperscale DCs
• The idea that satellite D2D can replace terrestrial mobile networks
• Hype that 5G-based (or, again, satellite) FWA fixed wireless access can replace massmarket fibre broadband (FTTH)
• Delusions that 5G can replace Wi-Fi for indoor or local nework connecivity (or, more rarely, vice-versa)

In reality, a system’s performance, risk, economics, controllability and governance will move through different “regimes”, which alter the way in which scaling works.

Scale curves are never smooth.

And while I initially focused on "substitutional scale" - tech X will replace tech Y, as it’s got an advantage with variable A - that is not the only way this error-class manifests. It’s actually a family of error types, which manifest in a variety of ways for whole industries or individual companies or products:

• Scale up organically - more of the same, growing a company or industry linearly or exponentially without thinking through the scaling regimes
• Scale up and substitute - as described here about space DCs, where people focus on a nice easy replacement / substitution narrative without realising all the aspects of inertia and power of incumbency.
• Scale down organically - the same, but for niches or "filling in the gaps". Taking a technology and assuming it works for SMEs, small deployments etc without thinking through frictions and fixed costs/needs.
• Scale down and substitute - fragment / decentralise something large. Sounds good, but it pushes coordination, security and governance to "the edge" or grass roots.

The full article explores all of these, gives examples, and explains why Scale Extrapolation Errors are so seductive, even for the smartest people in technology. It also discusses how to spot these errors - which should be top-of-mind for industry strategists, investors and policymakers.

Some other examples:

• A telecom operator assuming nationwide 5G rollout is just “more sites,” then discovering that site permitting, backhaul availability, indoor coverage, and energy costs start to dominate beyond certain thresholds.
• A “lightweight 5G core” for private networks that still inherits configuration and security complexity - and MNO-grade pricing
• Blockchain-based cellular or Wi-Fi networks aiming to compete with commercial-scale ISPs or MNOs, by incentivising individual islands of deployment.
• A “mini cloud region in a box” that still needs patching, monitoring, security updates, spare parts, and skilled operators.
• LLMs and agentic AI replace software: Improved variable: flexibility. Variables ignored: determinism, compliance, auditability, liability, predictable costs, integration, testing.

The full article is available for free here on my Substack. Please click the link and subscribe for additional updates and articles, such as tech & policy document critiques.

#telecoms #cloud #mobile #5G #6G #policy #regulation #AI #datacentres #FTTH #strategy