Where AI Value Lands

sme-connectivity-divide

3 min read

SME connectivity divide

Before any AI can be adopted, the firm needs the bandwidth and latency to use it. Across the OECD this baseline is gradually improving — but two persistent divides remain: a firm-size divide (large firms have high-speed broadband at far higher rates than small firms) and an urban-rural divide (metropolitan regions far outpace remote ones). Both shape who can actually participate in AI diffusion.

The firm-size divide

From OECD ICT Access and Usage data (Figure 8):

  • 2020: Large firms 69.4% / Small firms 41.9% — gap 27.5 pp
  • 2024: Large firms 87.6% / Small firms 60.0% — gap 27.6 pp

Adoption has risen on both sides — but the gap is essentially constant at around 25–28 percentage points across five years. This is a critical empirical finding: AI-readiness infrastructure is diffusing, but not converging. The relative position of small firms is not improving.

Why this matters: AI systems are integrated infrastructure, not isolated tools. They sit on top of existing IT stacks, cloud connections, data pipelines. Schwaeke et al. (2025) find that SME AI readiness is “intricately linked to their IT and technological infrastructures”. A firm without reliable high-speed broadband perceives AI investment as excessive — and is correct to.

The urban-rural divide

OECD’s 2025 connectivity study (covering 60+ countries with sub-national data):

  • Median fixed broadband speeds tripled 2019→2024 (53 → 178 Mbps).
  • But by Q4 2024, metropolitan fixed download speeds were 44% higher than in regions far from urban centres.
  • Mobile download-speed absolute gaps between metropolitan and remote regions widened from 5 Mbps to 45 Mbps over the same window — gap-in-levels growing nine-fold as the technology frontier advances.

Universal broadband policies are economically consequential for rural SMEs — broadband adoption raises firm scale and trade propensity (Kneller & Timmis 2016; DeStefano et al. 2022), particularly in knowledge-intensive sectors. Without intervention, rural SMEs are increasingly excluded from the productivity frontier.

G7 fibre and FWA pattern

  • Japan: 79% fibre share of fixed broadband — leads G7.
  • France: 70% fibre — second.
  • Germany: 12.2% fibre — far below OECD average; a notable structural lag.
  • US, UK: rapid Fixed Wireless Access growth (+39%, +30.4% y/y) — substituting for fibre coverage.
  • Canada: FWA at 7% of fixed broadband — appropriate for sparse rural geography.

The fibre lag in Germany is striking given its industrial-policy ambition for the Mittelstand — see country-profile patches.

Connectivity as the floor of the U

This divide places a hard floor under H2_u-curve-of-value: even if there were value to capture at the application layer, firms without high-speed connectivity cannot enter the layer in the first place. The “bottom” of the U is gated by infrastructure that is in turn gated by geography and firm size.

Also a direct concrete instance of diffusion-vs-innovation: the country that wins the diffusion marathon is the country that drags its connectivity floor up across firm sizes and regions — not the country that simply has the best fibre in the capital.

Open question

The OECD data doesn’t show whether the constant 25-pp firm-size gap is a steady-state phenomenon (small firms don’t need top-end broadband) or an underinvestment phenomenon (small firms can’t afford it). The implication for AI diffusion differs sharply between the two readings.

Graph View

Backlinks