GlyphAI

The grid has voted. The build-out hasn't stopped.

Ireland · UK · Grid capacity
The scarce input is no longer the chip.
It is the connection.

An Irish Examiner editorial points to a hard political fact: in an Ipsos B&A poll, 61% of Irish voters said data-centre construction should stop for now. Behind the mood are hard numbers. Data centres used 7,663 GWh in 2025, 23% of Ireland's metered electricity. In AI infrastructure, the scarce input is no longer the chip. It is the grid connection.

An Irish Examiner editorial (27 October 2026) cites an Ipsos B&A poll for the Irish Times: 61% of voters want data-centre construction stopped “for the time being at least”; 23% disagree. Behind the mood are the CSO's figures: data centres used 7,663 GWh in 2025, 23% of metered electricity, up from 5% in 2015. Their use grew 10% in a year, while total metered use grew 3.4%.
61%
Irish voters for pausing construction (Ipsos B&A)
7,663 GWh
Irish data centres, 2025 (23% of metered use, CSO)
125 GW
GB demand connection applications (≥80 GW data centres, Ofgem)

First two figures: Ireland. Third figure: Great Britain, Ofgem, July 2026.

The Local Cost & The Pricing Failure

The local cost is concrete. An Coimisiún Pleanála has overturned permission for the €1.5bn data centre in Killala, Co Mayo: about 300 construction jobs, but only 30–35 permanent roles. Ireland now lets new data centres and expansions proceed only if they bring their own power: on-site generation, or contracts for new, largely renewable supply nearby.

This is not an anti-technology mood. It is a pricing failure. We budget GPUs in depreciation cycles of a few years and socialise their power bill over decades. Residual value means little when the grid connection and community consent are both borrowed. Reliability is a social contract before it is an uptime SLA.

Case Study: 23,040 GPUs. A Grid Date in the 2030s.

Microsoft has reportedly withdrawn from Nscale's planned AI data centre in Loughton, Essex, first reported by The Telegraph. Microsoft did not comment; Nscale says the project is going ahead. The site was designed for 50 MW, expandable to 90 MW, with 23,040 Nvidia GB300 GPUs and first chips expected in early 2027. According to The Guardian, UK Power Networks warned that the required capacity might not be available until the 2030s.

My view: in AI infrastructure, the scarce input is no longer the chip. It is the connection. Silicon depreciates on a calendar. Grid capacity arrives on its own schedule. A GPU waiting years for a socket is not infrastructure. It is inventory.

Ofgem reports that demand connection applications in Great Britain rose from 41 GW to 125 GW in under a year, with data centres accounting for at least 80 GW. A queue that size does not show what will be built. It shows what has been promised.

GPU capacity plans in Ireland and the UK should start from the energisation date, not the hardware delivery date. Today, the grid connection is the bottleneck.

An H100 hour now costs 44% more than it did ten months ago.

GPU market · H100 The market prices the hour. Not the card behind it. H100 rental prices are up 44% since last December, for a generation that would normally be getting cheaper. Rental markets quote a model and an hour. As Hopper cards move into the resale market, the price will also have to reflect what each card has been through.

GlyphAI: Framework Analysis & Core Evaluation

GlyphAI · Framework review What is tested. What is still open. GlyphAI proposes a different unit for machine processing: not the words of a text, but its meaning, written as glyphs from a governed codebook. This is an honest look at where the framework stands today: what it claims, what has been tested, and what is still open.