Prepared for
PG&EPacific Gas and Electric Company
by Reddenda · TwinFlame Group

Ten gigawatts are asking to connect.

We call every customer. We hear the answer. You get capacity you can count.

▶ Listen first The founder briefing · 5 minutes Press play. By the end you will know exactly why this hour is worth taking. David Thomas, founder, on the ten gigawatts, the transformer wall, and the layer nobody has built.

Speed to power is now the product. Reddenda builds the two layers that decide how fast you can say yes: ground truth on where the load is going, and a conversation with every customer who can make room.

See the demand case → Hear a live call →
  • 10 GWdata center pipeline
  • 5.5 Melectric accounts
  • 70,000 sq miEureka to Bakersfield
GREATER BAY AREA · LOAD VIEW STANDBY
SANTA ROSA VACAVILLE STOCKTON CONCORD RICHMOND SAN FRANCISCO OAKLAND SAN RAMON TRACY CORRIDOR LIVERMORE FREMONT PALO ALTO SAN JOSE PACIFIC OCEAN S.F. BAY
~16Mpeople served
1,000miunderground, Oct 2025
65%EPSS ignition drop
7.5Mhomes of equivalent demand

Real Bay Area coastline, drawn from public state boundary data. Load concentration markers are illustrative: PG&E describes the data center pipeline as concentrated in the Bay Area, with the full territory running 70,000 square miles, Eureka to Bakersfield. Figures are PG&E's own published numbers, sourced at the foot of this page.

WHAT WE DO · IN ONE BREATH

We call every customer.We hear the answer.You get capacity you can count.

Five and a half million of them. In minutes. In their language.Yes, no, or silence. Confirmed, per person, graded per event.Faster connections. Smarter steel. Safer nights. All from one layer.

What this is worth to PG&E

The fastest headroom you ownFlexibility you can confirm per customer turns a years-long build into a yes on today's wires.
Every gigawatt served is about 1% off every billYour own math. Speed to power is affordability policy.
The one question no system asksCan you get out on your own. Asked, heard, and handed to the people who can act.
  1. The load you cannot seeTraffic becoming charging demand, feeder by feeder01
  2. The queue that leaks revenueReal projects stuck behind speculative ones02
  3. The capacity you already ownHeadroom stranded under a worst-hour rating03
  4. The customers nobody asks5.5 million relationships, zero conversations04
  5. The answer nobody confirmsReceipt is tracked. Capability never is.05

Demand growth · the load case

Two demand curves are arriving at once,
and only one of them filed an application.

Curve one · declared

The data centers, concentrated in the greater Bay Area

These loads announce themselves. They file, they site, they wait. They are enormous, they are clustered on a handful of substations, and they are the single largest new revenue event in your territory. The risk is not that they fail to arrive. The risk is that they arrive somewhere else because the wait was shorter.

Pipeline in queue10 GW
In final engineering, 17 projects1.5 GW

PG&E, July 31 2025. Expected to begin operations 2026 to 2030. PG&E puts the pipeline at roughly the demand of 7.5 million homes.

Curve two · undeclared

Traffic, becoming charging load, one transformer at a time

Nobody files an interconnection application to buy an electric car. The load simply appears, clustered by where people live and when they drive, on distribution equipment sized decades ago. Traffic patterns are the leading indicator of that load, and they are observable today, years before the transformer complains.

Reddenda already observes the road. Our transportation layer derives continuous, aggregate, privacy-clean vehicle counts and class mix from public cameras and open feeds. Pointed at your service territory it becomes a distribution-level charging forecast. No one is fusing these two things today.

One curve you can plan for because it filed. One you cannot see because it never will. Both land on the same feeders.

The metric that decides it

Speed to power.

Every load in your queue scores you on one number: application to energized. Copper does not set it. Intelligence does: which projects are real, where headroom already exists, and who will flex.

  1. 01

    Is the project real

    Queues fill with duplicates, options, and site-shopping. Capacity gets reserved by projects that never build.

    Reddenda: continuous, graded intent scoring on every applicant, updated as the world changes
  2. 02

    Where is the headroom

    Headroom is judged on a worst-hour static rating, so real capacity sits unused for all but a handful of hours.

    Reddenda: observed hour-by-hour load truth per feeder, so headroom is measured, not assumed
  3. 03

    Will they flex

    A load that agrees to curtail in the rare peak hour can often be served on today's wires. Almost no one asks.

    Reddenda: the conversation layer, at scale, with a signed and confirmed answer per customer
  4. 04

    Who else can make room

    Every battery, backup generator, fleet depot, and pump in the surrounding circuits is unmeasured optionality.

    Reddenda: enroll and confirm by voice across the whole surrounding class, in days
Credit where due · you have started

Flex Connect is real and it works: a fleet depot connected 18 months earlier under a peak-hour limit; charging sites accepted at a quarter of requested power at peak. T-Flex is announced for transmission. What none of it has yet: the per-customer layer that finds, confirms, and re-confirms the flexibility itself, at the scale of a service territory. That layer is what we build.

Exhibit A · we read the tariff

PG&E's interim large-load tariff, Electric Rule 30 per CPUC D.25-07-039, scanned in full.

"flexible"0
"curtailable"0
"non-firm"0

The only lever for a faster yes in the tariff is the customer's checkbook. The customer's load is not in the document. That is the opening.

Exhibit B · we read the alert flow

PG&E's own instruction to its most vulnerable customers, verbatim:

"Please acknowledge the alert. It is important to answer phone alerts or reply to text alerts so we know you have received them."

Receipt is tracked. Capability is not. No system of record answers the one question that matters in an evacuation: can this person get out on their own.

The capacity you already own

Ask the wire what it can carry today,
instead of what it could carry on the worst day of the year.

A static rating is one number chosen for the worst hour, applied to every hour of the year. Real capacity moves with weather and flow. The gap is headroom you already paid for: no permit, no steel.

CONDUCTOR CAPACITY · ILLUSTRATIVE STATIC vs COMPUTED
COMPUTED CAPACITY STATIC RATING ACTUAL FLOW

Illustrative shape, not measured data. The point is structural: a static rating is a flat line chosen for the worst hour, and the real limit is a curve that spends most of the year above it.

Compute the rating, do not assume it

Weather, conductor condition, and sensing resolved into a live capacity figure per span, with the confidence interval attached. The number an engineer can put in a filing.

Find the constraint that actually binds

A circuit is limited by one element for a handful of hours. Knowing which element, and when, turns a rebuild into a targeted fix, or into a flexibility agreement instead.

Turn headroom into a yes

Once real capacity is measured hour by hour, a large load that agrees to stand down in the binding hours can often be served on today's wires. That is speed to power without a build.

Grade every call against what happened

Each capacity call is logged and scored against the realized outcome. The forecast earns its confidence over seasons, which is the part a competitor cannot backfill.

Speed to power, continued · intelligence, coordination, grid status

The grid knows what it needs.
The customer has never been told.

Your system knows which circuits are tight. That picture stops at your fence. On the other side sit 5.5 million customers who could move their own load, if anyone asked. Coordination is the cheapest capacity in the system, and it is almost entirely unbuilt.

Grid status

A live, shareable picture of what the system needs

Not a public alert. A per circuit, per hour statement of the constraint, with the window and the size of the ask.

Coordination

The ask goes out as a conversation, not a broadcast

Every customer who could help hears a real question in their own language, at a moment that still leaves them a choice, and answers in their own words. The answer comes back structured.

Customer levers

What a customer can actually move, once they are asked

  • Delay vehicle chargingHours of flexible load in nearly every residential circuit
  • Pre-cool or pre-heatThermal mass carries a building through the binding hour
  • Discharge a batteryStorage already sited behind the meter, mostly idle at peak
  • Run backup generationCapacity you never built, sitting in yards and basements
  • Shift a process or a shiftPumps, irrigation, cold storage, industrial batches
  • Stand down on requestThe large load that agrees to yield in the rare binding hour
The result

A number your planners can actually count on

Not an estimated participation rate. A list of named customers, each with a confirmed commitment, a size, and a time window, refreshed before every event and graded after it. That is what turns coordination into capacity, and capacity into a faster yes.

The arithmetic, independently published
4.2 to 5.9GWof headroom in CAISO alone if new large loads curtail 0.25 to 1 percent of annual hours, per Duke University's 2025 headroom study
1.7hrsthe average curtailment event at the smallest case. Most hours retain at least half the load. This is a schedule adjustment, not an outage
180hrs/yrwhat PG&E's existing Base Interruptible Program already tolerates. The bargain is proven. What is missing is the layer that asks and confirms at scale

Internal operations · the transformer wall

287,000 new transformers by 2040.
Or know, per unit, which ones can wait.

Your own study says the install rate has to nearly double for fifteen years, while lead times run two to five. No supply chain rescues that. Triage does: which unit is truly next, which can wait, and which can be bought years by the homes beneath it.

DISTRIBUTION TRANSFORMERS · TRIAGE VIEWILLUSTRATIVE
Healthy Can wait, flexibility confirmed Replace next

Illustrative triage wall. The real one ranks every unit by observed load, incoming charging demand, and the confirmed flexibility of the customers beneath it.

74,500transformers overloaded by 2040 without intervention, base scenario
10,841 → 19,966installs per year: the historical rate vs what the forecast now requires
$3.4Bof secondary-system cost driven by overloaded service transformers through 2040
5 to 15yrsof deferral per feeder when charging is managed, worth over $300,000 per feeder
01

See the cluster forming

EV adoption clusters block by block and breaks system-level forecasts. Observed traffic and charging behavior shows the cluster years before the overload.

02

Ask the homes beneath it

A confirmed charging-window agreement from the households on one transformer is deferral you can bank, unit by unit, not a program average.

03

Spend the fleet where it counts

Every install goes to a unit that truly cannot wait. Every deferral is documented, confirmed, and graded against what the transformer actually did.

Figures from the 2026 PG&E Electrification Impact Study, prepared by E3, and PNNL's charge-management deferral analysis. Linked in the sources.

Commercial and residential · in your own words

Every confirmed customer answer
is a rate case argument.

1 GW → ~1% PG&E's own statement: under the right conditions, one gigawatt of new load can reduce rates by about one percent. Serving the queue faster is affordability policy.
80% of applications in PG&E's latest cluster study moved to areas with available capacity after outreach. Steering works. It works better when the steering layer can hold a real conversation.
182 days the CPUC's SB 410 average energization target. Every layer on this page exists to make that number beatable, and to document why when it is not.
Exhibit C · we read the post-event filings

How a "successful positive notification" is defined for the customers who depend on power to live, verbatim from PG&E's PSPS post-event reporting, unchanged from the January 2025 filing to the July 2026 filing:

"If a notification is acknowledged at any stage and/or a door hanger is left, that is considered a successful positive notification."
in scope288
attempts288
"successful"288

A one hundred percent success rate, under a definition where a hanger on an unanswered door counts. Nobody is lying. The measurement just stops at the door. Our layer is what stands on the other side of it: an answer, from the person, or their absence surfaced as a fact an incident commander can act on. The evacuation offer, in full →

What Reddenda offers PG&E

Four layers. One capability underneath them.

5.5 million relationships. Zero conversations. Everything below is what becomes possible the moment you can have them.

Live today

Confirmed demand response

Before a peak or a constraint, call the relevant class of customers and ask a real question. Get back a graded, per-customer answer: who will shift, who will curtail, who cannot. Not an opt-in rate assumption. A confirmed commitment with a name on it.

  • Enrollment that closes on the call, not in a mailer
  • Availability confirmed hours before you need it
  • Every answer logged with its timestamp and provenance

Today: enrollment is an opt-in program and an event signal. Participation is estimated afterward, never confirmed before.

What changes: peak capacity you can commit, because it answered.

Live today

The generation you already have

Batteries, backup generators, solar, fleet depots, agricultural pumps. Your territory is full of dispatchable capability you did not build and cannot see. The barrier to using it has never been technology. It has been that nobody can reach every owner in time.

  • Enroll and re-confirm availability across a whole class in days
  • Verify a resource is actually there before you count on it
  • Turn scattered assets into a number an operator can plan against

Today: dispatch is fire-and-forget, settled after the fact. No program confirms a device is actually available before the event.

What changes: a fleet you never built, counted because it is confirmed.

In development

Load truth per feeder

Headroom judged on a static worst-hour rating leaves real capacity stranded. Observed, hour-by-hour load truth, fused with the traffic that is about to become charging demand, turns headroom from an assumption into a measurement you can defend in a filing.

  • Where the next constraint forms, before it forms
  • Which feeders can take a large load today
  • Every forecast graded against what actually happened

Today: EV load is discovered by reading the meter after chargers energize. Nobody joins observed traffic to a feeder name.

What changes: the constraint seen years early, the upgrade placed once, correctly.

Live today

The evacuation nobody confirms

EPSS and undergrounding have driven ignitions down. The remaining exposure is the hours around a de-energization or an evacuation, when every system you own can send a message and none of them can hear an answer. We call every home and ask the one question a text cannot.

  • Two lists back: who needs help, who was never reached
  • Contact attempts documented per customer, per event
  • Built for the customers a notification is most likely to fail

Today: the vulnerability list and the evacuation authority sit in two different organizations, and the alert confirms receipt, not capability.

What changes: two lists an incident commander can work down, per event.

See the full DoorknockAI offer for PG&E →
The capability underneath

A voice layer that has a real conversation with a person, at the scale of a service territory.

It speaks the customer's language, it hears the answer, it grades the answer, and it hands your team a structured list instead of a delivery report. It is running today. You can hear it in ninety seconds.

Hear the call →

What we will and will not claim

The line is the architecture, not the disclaimer.

  • Reddenda never trips a breaker, sheds load, or actuates a device. Every output is a warning or a confirmed answer routed to a human. A person decides and acts.
  • The voice layer is live. The grid intelligence layer is in development. Zero production utility deployments, no certified system. We will not pretend otherwise on a page meant to start a real conversation.
  • No guarantee of uptime, outage prevention, or a specific interconnection timeline. We sell earlier information and a confirmed answer, not an outcome we do not control.
  • We watch assets, endpoints, and aggregate movement. Never persons. The traffic layer derives aggregate counts only, and identifies no vehicle, plate, face, or person.
  • The pieces exist as products, and you already buy some of them. Managed charging exists. Telemetered VPPs exist. Dynamic line ratings exist. Notification platforms exist. What we could not find anywhere is the join: observed traffic fused to a feeder name, a spoken answer confirmed per customer and graded per event, and a flexibility commitment written into a large-load yes. Where a device has telemetry, we do not replace it. We reach the millions of customers who will never install one.
  • Every figure on this page is PG&E's own published number or a cited public source. Anything we could not verify from a primary source is not on this page.
One layer. Both ends of the wire.

The grid side predicts the failure.The human side reaches the person.

Hear it from David ThomasHow the lights actually go out, and who is on the other end · 6:32
Nobody else does both. Most do neither.
01TuendaSees the pole about to fail weeks before it does. No more surprise outages.The grid side 02CAgridEvery asset, every risk, one screen. The picture no utility has ever had.The operating picture 03DoorknockAICalls every home and asks if they can get out. Finds the ones who can't.The human side

The ask

One working session.
Bring the queue.

Ninety minutes with your large-load, distribution planning, and customer operations people. We will take one real constrained substation and one real cohort of customers, and show you exactly what the two layers would have told you, and what you could have said yes to.

Book the working session → The evacuation offer for PG&E → See the grid layer

David Thomas · Chief Executive · Reddenda · TwinFlame Group