Issue 21: Power Limits Slow Central Buildout
Permanent copy of issue 21. Today's issue is always the front page.
Central clusters keep ordering chips and power, but open laptop-scale weights and local runtimes now divert measurable work from paid APIs.
Date: 2026-08-24
Updated: 2026-08-24 02:00
Status: LIVE
Broadcast
Central spend keeps rising while power queues lengthen. Hyperscalers now forecast $3.6 trillion in AI capex through 2028, yet Texas has stopped new grid approvals for data centers. At the same time Alibaba released a laptop-scale open model meant to run locally. Token list prices continue to fall, and on-device runtimes capture more developer workloads. The net effect is slower central growth offset by measurable local substitution.
Watch
Stories
Hyperscalers spend 102% of cloud revenue on AI capex
Yahoo Finance reports hyperscalers are spending more on capital expenditure than they collect in cloud revenue. The gap shows central buildout continues at full pace.
Central operators keep ordering chips and power faster than revenue grows, locking spend into large clusters.
2026-08-22 / Yahoo Finance / chip, energy / central
BofA lifts hyperscaler AI capex forecast to $3.6 tn through 2028
Bank of America raises its five-year AI infrastructure outlook. The revision signals sustained central demand for chips and grid capacity.
Larger central budgets lock more work into hyperscale clusters rather than local hardware.
2026-08-18 / Investing.com / chip / central
Alibaba ships laptop-scale open-weight model to rival Meta
Alibaba releases a new open model sized for consumer laptops. The release gives developers a local alternative to hosted APIs.
Laptop-ready weights move inference away from paid central APIs toward on-device execution.
2026-08-17 / CNBC / efficiency / local
Texas halts new data-center grid connections
State regulators stop approving fresh data-center links to the grid. Power queues now slow the pace of central cluster expansion.
Grid constraints raise the cost and delay of adding more central capacity, tilting marginal work toward local hardware.
2026-08-19 / Forbes / energy / central
Axis
local [###################-----------] central
Score: 62
clusters still winning
-1 toward local this week
Most AI still runs in big central data centers, but open-weight models and on-device runtimes are taking some work off paid APIs.
Forces
| Force | Meter | Score | Pull | Note | ||
| CHIP / CAPEX |
|
74 | central | Hyperscaler AI capex forecasts now exceed $1 trillion for 2026 with $2.3 trillion in backlogs. | ||
| ENERGY |
|
82 | central | Texas halts new data-center grid connections; power queues now limit buildout speed. | ||
| EFFICIENCY |
|
41 | local | List token prices keep falling; on-device runtimes keep eating rented work. |
Line charts (0=local 100=central 2026-08-09 to 2026-08-24)
Market overlay (2026-07-26 to 2026-08-24)
Latest: BTC $77,183 / hashrate ~887.93 EH/s / token markers avg $5.3033/M / axis 62. These lines are different units. We compare timing, not dollars-to-hashrate.
Not shown: GPU cloud $/hour, hub electricity, miner hashprice, or lab cost per token -- none of those are a clean public feed.
Power and list prices
BTC $77,183 / hashrate ~888 EH/s. List-price gap (not a subsidy %): 44.4x list gap (blended) GPT-5.6 Luna -> Claude Fable 5 ($0.45/M to $20.00/M). (mempool.space + Coinbase/CoinGecko)
Marker prices (USD per 1M tokens; verified 2026-08-19; blended = (3*input + 1*output) / 4; -- = open/self-host or no single public list)
| # | Model | Org | CC | Weights | Tier | In | Out | Blend |
| 1 | Claude Sonnet 5 | Anthropic | US | closed undisclosed |
workhorse | $2.00 | $10.00 | $4.00 |
| 2 | GPT-5.5 | OpenAI | US | closed undisclosed |
flagship | $5.00 | $30.00 | $11.25 |
| 3 | DeepSeek V4 Flash | DeepSeek | CN | open ~284B total / ~13B active (MoE) |
commodity | $0.44 | $1.32 | $0.66 |
If useful, support the broadcast.
bc1qfs3yw5qlq8sxs50crzh3g84ug27gvww9npu85n
home | rss | archive | models | data | board | mcp | llms.txt