In the previous article "How AIoT Firms Shift to Outcome-based Profit Sharing & Escape Project-based Cash Flow Dilemma", I explored a proposition: how can IoT companies break free from the fate of merely selling hardware and engaging in price wars, and share profits based on the business results created by AI? The core conclusion of the article is that to get results, we need to bridge the gap between measurability and attributability.
At industrial sites, the most objective metering method is to directly read the electricity meter, because behind the meter lies the credibility supported by national standards and grid standards.
In fact, not only electricity, but also water, gas, heat... various types of continuously flowing physical data, once equipped with recognized metering specifications, have the opportunity to be priced by output just like Tokens.
A recently implemented group standard happens to provide a model for this "assetization of physical quantities".
Next, taking this standard as an example, we will analyze how IoT companies can step by step transform "physical quantities" into "settlement vouchers".
In September this year, the Shenzhen IoT Industry Association released four group standards for virtual power plants, which came into effect on October 1. A virtual power plant has no physical plant; it aggregates distributed photovoltaics, energy storage, air conditioning, and charging piles, adjusting power generation and consumption according to grid commands, participating in regulation just like a real power plant.
Among them, the "Technical Specification for Measurement Devices of Virtual Power Plants" (T/SZIOT 031—2026) has a key regulation: in addition to factory qualification, the measurement device must undergo strict on-site acceptance and be equipped with a custom seal from the grid's virtual power plant management center.
This acceptance seal is profound. It is no longer a simple hardware quality inspection. Viewed through the business logic of the AI era, this seal is essentially a "Token minting machine" in the physical world.
It means that the readings of this meter have been trusted by the grid, and every load adjustment has a metering basis for participating in settlement.
Today, AI companies mostly charge for model calls based on the measurement unit of Tokens, with clear bills; while IoT companies, after installing sensors for over a decade, still struggle to issue a bill priced by output.
The reason lies in the fact that over the past decade, we have only been responsible for manufacturing devices and connecting them to the network, but have failed to turn the collected data into settlement units with transaction consensus like Tokens.
My view is that for IoT companies to build their own "Token factories", it is not far from issuing this output-priced bill. What we really need to bridge is to make up for the following three core elements:
A set of value accounting baselines recognized by all parties;
A dispatchable asset that can serve two buyers simultaneously;
And a trusted metering terminal that transforms physical quantities into commercial Tokens.
The First Step: Establishing a Consensus Baseline to Make "Estimated Output" as Countable as Tokens
In the AI era, data centers are essentially "AI factories", and their core product is called Tokens. A Token is the smallest measurement unit for large models to read and write text; calls are measured and priced by it.
This business is very clear: output multiplied by unit price is revenue. AI companies can charge by output, relying on three conditions:
1. Output is absolutely countable: one Token is one Token, what you see is what you get, with no gray areas.
2. Clear asset ownership: the factory holds and operates the equipment, so the output naturally belongs to it.
3. Universally recognized and undisputed metering: the metering is trusted by the buyer, the bill is issued by the seller, the buyer pays according to the bill, and the generation record of Tokens is certain in the digital world.
In contrast, most IoT companies lack all three conditions. We have fallen into an awkward mismatch: the number of connections sold and the online rate, although countable, are not what customers are willing to pay more for; while the "energy saving" and "loss reduction" that customers are truly willing to pay for cannot be counted.
Why can't it be counted? Because energy savings is a special kind of "estimated output". It equals "how much electricity would have been used" minus "how much electricity was actually used".
Actual power consumption can be read from meters, but "how much would have been used" is an unknown that cannot be directly measured. We must first estimate a reasonable reference value to calculate the quantity, and this reference value is the "baseline".
If the baseline is not set accurately, pricing by output is just empty talk.
In the field of virtual power plants, the "adjustment volume" sold is also an estimated output, and it is necessary to know how much electricity would have been used in the first place. In the past, the assessment and settlement of virtual power plants lacked unified and quantifiable indicators, with different calculation calibers in different regions. Evaluating the same virtual power plant in a different place might yield different results.
The newly released "Design and Calculation Methods for Adjustable Performance Indicators of Virtual Power Plants" (T/SZIOT 032—2026) hits the pain point directly. The standard lists 17 general indicators and 6 specific scenario indicators, providing clear algorithms for each item, with the most critical being the baseline.
For example, the baseline power takes the average power of the most recent un-called 15 minutes before the call; the baseline load generally takes the corresponding time periods of the most recent 5 similar days without response, and calculates the average after excluding outliers. "How much electricity would have been used" now has a unified mathematical formula in the field of virtual power plants that anyone can recalculate.
I once proposed the "Hundred Billion Connection Trap". The industry talks about growth in terms of connection numbers; connection numbers rise, but revenue does not keep up. The root cause lies in the mismatch between "countable" and "valuable". What the 032 standard does is to unify the caliber of the algorithm in the middle.
Of course, current group standards and national standards are mostly recommended, and settlement rules still vary by region. With the algorithm unified, baseline disputes will not disappear instantly, but the industry finally has a common starting point. Establishing a consensus baseline means IoT companies have truly taken the first step in building a "Token factory".
The Second Step: Mastering Asset Dispatch Rights to Issue a "Second Bill" for the Same Set of Equipment
The previous article mentioned that Mogu IoT observed a trend in practice: after AI enters industrial sites, compared to finding engineering companies to retrofit equipment themselves, customers are more inclined to adopt BOT (Build-Operate-Transfer) and BOO (Build-Own-Operate) models for cooperation, although most customers still choose one-time buyouts at present. Such models can be collectively referred to as "energy hosting".
Following the logic of energy hosting, we can infer a deeper commercial value: after equipment hosting, how the air compressors are started and how much load they carry are dispatched by the IoT operator under the premise of meeting the factory's gas demand; and what virtual power plants need are precisely such high-quality resources that can be dispatched according to commands.
Thus, the same air compressor station may issue two bills:
The first bill: collected from the factory based on output (earning money from energy hosting);
The second bill: collected from the electricity market via the virtual power plant based on the adjustment volume (earning money from grid regulation).
Regarding this "second bill", there are actually precedents in the computing power industry.
Data centers not only make money by selling computing power services but also by participating in electricity spot trading.
For example, the three large data center clusters of China Unicom in Shaoguan, and China Mobile in Guangzhou and Zhanjiang, once participated in electricity spot trading through the virtual power plant platform of Guangdong Power Grid Energy Investment Company. They increased computing power tasks (using more electricity) during the low electricity price period from 10:00 to 14:00, and decreased tasks (using less electricity) during the high electricity price period from 16:00 to 20:00, thereby obtaining a spot price difference revenue. This is a typical "second bill".
Then, how should IoT companies sign this second bill?
The newly released group standards provide the path.
The "General Requirements for Technical Support Systems of Virtual Power Plant Operators" (T/SZIOT 033—2026) defines proxy users (resource owners) and requires the operator system to be able to manage proxy contracts and tripartite agreements.
Taking the air compressor station as an example, the equipment control right of the hosted station is in the hands of the IoT enterprise (the hosting party), but the electricity consumption account number is usually under the name of the factory (the owner). Therefore, this bill requires at least the owner, the hosting party, and the virtual power plant operator to sign.
Understood in business language: the factory can also bypass the hosting party and sign directly with the virtual power plant operator; for an IoT enterprise to secure a seat in the second bill, the premise is to first win the first bill (obtain equipment dispatch rights).
The power consumption scale of a single station is limited, and it usually needs to be aggregated by the operator to cross the access threshold of the local grid regulation authority.
Faced with this incremental revenue, IoT companies have two paths to choose from: either enter the field themselves to become virtual power plant operators, at the cost of having to make up for the core capabilities of power trading and dispatch; or act as one party in tripartite cooperation, using the "equipment dispatch rights" in hand as a bargaining chip to write the profit-sharing terms into the agreement.
The Third Step: Implementing Effective Control to Complete the Trust Loop for Physical World Output
This final step is divided into two halves, front and back.
The first half is moving from "connectable" to "controllable".
For any resource-aggregation business (whether computing power networks or virtual power plants), monetization at scale must pass four checkpoints: connectable, controllable, callable, and settled.
A set of data from Shenzhen reveals the harshness of the first two checkpoints: the power supply bureau platform has accessed over 60,000 resources with a scale of over 5.1 million kilowatts, but the maximum adjustable capacity is only slightly over 1.4 million kilowatts.
Roughly calculated, in terms of capacity, less than 30% of the accessed scale can be dispatched according to commands; a single load can only be adjusted by a part under the premise of not affecting production, and part of this gap is the physical limit. We might as well call the proportion transformed from "physical access" to "commercial availability" the "effective control rate".
The profit of a Token factory depends on the utilization rate; idle GPUs do not generate Tokens; similarly, capacity hung on the platform but unable to be dispatched cannot get response revenue.
When the "Hundred Billion Connection Trap" falls into specific commercial monetization, it means the paper scale shrinks step by step at these four checkpoints.
The second half is making others trust your readings.
The output of a Token factory is in its own system, but the output of the physical world occurs in the customer's workshop and on external networks. There is an iron law in the business world: producers cannot be their own judges.
The 031 standard provides the answer to break the deadlock: the AC quantity collection accuracy of the measurement device must meet the requirements of Class C electricity meters. After the device is installed by the operator, a qualified third-party agency or a unit designated by the grid enterprise must go to the site to verify the wiring, compare the readings, confirm anti-theft measures, and finally install a custom seal from the grid enterprise's virtual power plant management center.
These four group standards placed together exactly correspond to the "four rights" in the business system:
The "Technical Specification for Grid-Connected Access of Virtual Power Plants" (T/SZIOT 030—2026) manages collection (reading right): it stipulates that data is collected according to electricity consumption account numbers, provides 101 measurement point codes, and terminal encryption keys are uniformly distributed by the grid company, defining who can get the data;
The adjustable performance indicator standard (032) manages the algorithm (pricing right): clarifies how the adjustment volume is calculated, defining what to calculate money by;
The measurement device specification (031) manages the meter (acceptance right): stipulates the seal mechanism, defining who confirms the work is done;
The operator technical support system standard (033) manages the system (bookkeeping right): requires daily clearing and monthly settlement for proxy users, defining who keeps the ledger recognized by all parties.
The qualification certificate when the hardware leaves the factory can only prove that the equipment itself has no problem; while this "custom seal" installed on-site truly proves that the underlying data has been trusted by the grid, which is the most core trust mechanism in business transactions.
For the second bill to collect real money from the market, the mark of IoT companies stepping onto the third step is to have their installed meters pass strict acceptance, successfully stamped with the exclusive seal of the grid enterprise, and make up for the trust links required for commercial monetization.
Conclusion
As mentioned before, the reason why electricity meters are the most objective is that behind them are national standards and grid enterprise standards.
But this sentence actually has a more core other half: any meter used for trade settlement must undergo mandatory verification and be sealed by others.
Token factories in the purely digital world can prove their output in their own systems, but IoT companies moving into deep waters need to hand over the trust loop to consensus. And the establishment of such business consensus usually relies on several clear paths:
A set of unified standards penetrating the underlying layer (using unified protocols and data dictionaries to define what to collect and how to read);
A physical verification by an authoritative third party (using on-site acceptance and custom seals to establish trust boundaries);
Or a traceable and auditable ledger (using algorithms and systems recognized by both parties to record settlement and profit sharing).