The Pulse™ | Week 29: The map is the moat
Five signals from the fog. Coverage window: Mon July 6 to July 10, 2026.
By Armando Pereira | Founder, PVentures Consulting | Senior Member IEEE | Co-founder, OpenFog Consortium (IEEE 1934) | President, Autonomous Vehicle Computing Consortium | Former VP/GM Optical BU, Centillium Communications (CMOS PON SoC, NTT-qualified)
👋 Welcome back to The Pulse™
The Pulse™ is your weekly operating brief for execs, founders, and investors. What changed across Industrial IoT, Telecommunications, Edge Computing, Autonomous Systems, and Artificial Intelligence, why it matters, and what to do next, all in one place.
🩺 This Week’s Pulse™
The signals this week point to the same shift. The edge that matters in deep tech is moving from what you build to where you sit,
A carrier that wins the right spectrum band,
A data center that lands at the right power interconnect,
A robotaxi that holds the right city, and
A factory that owns the silicon on its own floor.
All win the same way: by claiming a scarce physical position before rivals do. Technology still decides who qualifies. Position decides who profits. This week the map became the moat.
Share this: In 2026 deep tech, the moat is no longer the better technology; it is the scarce physical position that technology needs: spectrum, power, territory, and silicon on the floor.
🏭 INDUSTRIAL IOT: AMD puts factory intelligence on one board
AMD confirmed that its Ryzen AI Embedded P100 series enters production shipments this month, an industrial-grade processor that integrates programmable logic control, machine vision, and the operator interface into a single industrial PC. For plant managers, OT and IT convergence is arriving as hardware, not slideware.
Up to 80 TOPS of system performance for physical AI on one x86 module, sampling now and shipping in July.
One board replaces separate PLC, vision, and HMI boxes, cutting cabinet count and integration cost.
Fewer discrete controllers mean a smaller attack surface and simpler patching for OT security teams.
Expect a wave of single-PC line retrofits over the next two quarters as OEMs certify the part.
📡 TELECOMMUNICATIONS: A spectrum week decides the next decade
Regulators on three continents handed out positions this week. Japan’s ministry closed its 26 GHz auction, with NTT Docomo taking the nationwide license for 400 MHz of millimeter-wave spectrum, and the US FCC set a vote to auction 160 MHz of upper C-band, the opening move on a 440 MHz super-band aimed at deployment by 2030. I saw this pattern up close on the Marvell broadband program, where NTT Labs qualification, not marketing, decided who shipped.
Docomo wins nationwide 26 GHz; JTower and High Tech Inter take regional millimeter-wave licenses.
The FCC’s upper C-band vote frames the defining US spectrum story of the year.
Netherlands, Poland, and New Zealand open private-network bands, feeding the enterprise 5G pipeline.
Spectrum won now sets the cost base for every carrier and private-network build through 2030.
⚡ EDGE COMPUTING: On-device inference crosses into production
The edge story this week is economic, not a single launch. Fresh market data put the AI inference chip market at $20.5 billion in 2026, up more than 15%, as small language models move the bulk of enterprise inference onto local silicon. A seven-billion-parameter model served at the edge now runs 10 to 30 times cheaper than a large model in the cloud. This is the deployment substrate the OpenFog Consortium described when we authored the reference architecture that became IEEE 1934: compute placed where the data is created.
The inference-chip market is projected to reach $20.5 billion in 2026 and nearly $37 billion by 2030.
Small models, such as the Phi-4 family, now handle routine, latency-sensitive workloads on-device.
Hybrid routing, small models local and large models in the cloud, becomes the default enterprise pattern.
The question shifts from whether to run inference at the edge to which workloads stay local.
🚗 AUTONOMOUS SYSTEMS: Waymo claims four more cities
Waymo said on July 8 that it will begin fully driverless rides in San Diego, Las Vegas, Tampa, and Denver, starting with Alphabet employees before opening to the public. With roughly 4,000 vehicles already running fifth- and sixth-generation systems, the contest in robotaxi is now about territory and operations, not just perception. Governance scaffolding matters here too; the AVCC, which I led as founding president, existed to provide this compute stack with a common safety and architectural baseline.
Four new metros extend driverless service across a wider footprint in the Sun Belt and Mountain West.
California’s rules lifting the ban on driverless trucks over 10,001 pounds took effect July 1.
Rivals must now match Waymo on city coverage and reliability, not demos.
Whoever locks in cities and safety records first sets the terms competitors inherit.
🤖 ARTIFICIAL INTELLIGENCE: Power becomes the AI asset
The week’s biggest AI move was not a model. EQT agreed on July 10 to acquire Copia Power, whose model co-locates generation, high-voltage transmission, and data center load at a single interconnection point, with a 9 GW pipeline of grid-connected data centers. Microsoft separately stood up a $2.5 billion company to help enterprises deploy AI. The scarce input has moved from chips to deliverable megawatts at a real address.
EQT to buy Copia Power; the pipeline includes 9 GW of data centers plus 25 GW of solar and storage.
Co-locating power and compute at one interconnect shortens the timeline to live capacity.
Microsoft’s $2.5 billion deployment unit signals that services, not just models, are the bottleneck.
Board questions about AI now start with power procurement rather than model choice.
🔥 3 Non-Obvious Takeaways
1. Spectrum, power, and territory are now the same asset class.
The Docomo auction, the Copia power deal, and Waymo’s four-city expansion are three faces of one trade: pay early to lock a scarce physical position, then let competitors bid against your cost base. Carriers, hyperscalers, and AV operators are all in the same business of claiming interconnection points before rivals can.
2. On-device silicon is a factory security decision, not only a performance one.
AMD’s single-board industrial PC and the edge inference shift look like speed stories, but their sharper effect is on attack surface. Fewer discrete controllers and less cloud round-tripping shrink what an OT security team must defend. The plants that consolidate compute early will also be the plants that are easier to secure.
3. The cloud giants are quietly conceding the edge.
Google, Microsoft, and AWS keep building central capacity, yet this week’s momentum ran the other way: inference moving on-device, power moving to the interconnect, factory intelligence moving to the floor. The center still trains the models. The edge increasingly runs them.
🗺️ The arc so far
This week extends a thesis The Fog Signal™ has tracked since May: control is descending the stack, and this issue plants it in physical geography.
The Sextant™ (May 6): “The Hidden Bottleneck in AI Compute.” Named compute capacity as the binding constraint on the AI buildout.
The Vector™ (May 14): “The Inference ASIC Fork.” Traced the silicon split between training and inference economics.
The Sextant™ (May 20): “The Hidden Lock-In Beneath Inference and Physical AI.” Showed how that split hardens into vendor lock-in.
The Pulse™ (June 9): “The AI Stack Goes Institutional.” Governance and standards entered the stack.
The Pulse™ (July 7): “Breaking the lock-in.” Players spent capital to own their layer and route around vendors.
This issue, Week 29 (July 14): ownership becomes geographic; the moat is the physical position, spectrum, power, territory, and silicon on the floor.
🧭 Where to start
If claiming physical position, spectrum, power, territory, or on-device silicon is becoming a live decision for your leadership team, the Board Advisory Session is the right starting point. It is built for boards that need to pressure-test where their company sits on the map before a competitor sets the cost base:
❓ Question for you
Pick the vertical you own, whether AI, Telecom, Edge Computing, IIoT, or Autonomous Systems. Does your team have the instrumentation to detect when a scarce physical position in it is about to be claimed by a competitor?
If you are not tracking spectrum and interconnection timelines, on-device compute footprint, and territory or coverage expansion by rivals, you are making 2026 procurement calls without the leading indicators that separate early movers from laggards over the next two quarters. Position is bought before it is obvious; the dashboards that catch it early are the ones that pay for themselves.