PJM: Producer - Pay to Play

 

PJM adopts the common practice of new power producers paying for the grid upgrade required to support the new load being introduced by the producer. This ensures fairness as a principle and avoids free-riding. 

A similar model exists for most markets including MISO, ISO-NE, SPP, ERCOT, NYISO, although with e.g. ERCOT, broader upgrades beyond the interconnection can be shared with the transmission operator funded through transmission rates.

However: In PJM, MISO, ISO-NE and SPP, transmission owners also have the unilateral ability to fund upgrades and charge the interconnection owner for the cost plus a return. There are concerns by FERC that this increases the cost for end consumers because the transmission owners is earning a return for the upgrade vs. this being paid for by the generator - although still to be reconciled that funding of capex should in general earn a return to attract capital, regardless of the source of that capital. The transmission owners should therefore have the right to compete to fund upgrades. But should this be at the transmission owner or the producer’s discretion?

Another concern is around cost allocation - how much of the burden is funded by the producer, and balancing that against burdening the rate base to be shouldered by consumers who may not ultimately benefit from the upgrade.

Read More

Anatomy of an Auction: A Simplified Example of PJM Capacity Auction Clearing

PJM Service Area (inc. DC)

Anatomy of an Auction: A Simplified Example of PJM Capacity Auction Clearing

The PJM annual capacity auction utilizes a "descending clock" mechanism to determine which power generators will be paid to be on standby and ensure a reliable electricity supply for the 65 million people in its territory. While the actual auction is highly complex, this simplified example illustrates the core principle of how the auction "clears" and sets a uniform price for all successful participants.

The Setup: Demand and Supply

First, PJM determines the total amount of capacity it needs to secure for a future year. This is based on forecasted peak demand plus a reserve margin for unexpected events.

Let's assume for our example:

• PJM's Required Capacity: 10,000 Megawatts (MW)

Next, power plant owners offer their available capacity into the auction at the minimum price they are willing to accept to be available for that year. These offers come from a variety of sources with different costs.

Here are our hypothetical power plants and their offers, ranked from lowest to highest price:

Power Plant	Resource Type	Capacity Offered (MW)	Offer Price ($/MW-day)	Cumulative Capacity (MW)
Plant A	Nuclear	2,000	$10	2,000
Plant B	Hydroelectric	1,000	$25	3,000
Plant C	Coal	3,000	$50	6,000
Plant D	Natural Gas (Combined Cycle)	2,500	$90	8,500
Plant E	Natural Gas (Peaker)	2,000	$150	10,500
Plant F	Older Coal	1,500	$175	12,000
Plant G	Demand Response	500	$200	12,500

The Descending Clock in Action

The auctioneer starts the "clock" at a high price, well above what anyone has offered. At this high price, all the power plants are considered "in" the auction because the price is favorable to all of them.

The auctioneer then systematically lowers the price in rounds. As the price on the clock drops, it will eventually fall below the offer price of some of the more expensive power plants. When this happens, those plants will drop out of the auction as the price is no longer acceptable to them.

Let's simulate the process:

• Clock Price: $250/MW-day: All plants (A through G) are "in." The total available capacity is 12,500 MW, which is more than the 10,000 MW needed.

• Clock Price: $180/MW-day: The price is still above Plant F's offer, so it remains in. All plants from A to F are still in.

• Clock Price: $160/MW-day: The clock has now dropped below Plant F's offer of $175/MW-day. Plant F drops out. The total available capacity is now the sum of Plants A, B, C, D, and E, which is 10,500 MW. This is still above the 10,000 MW requirement.

• Clock Price: $150/MW-day: At this price, Plant E, with its offer of $150/MW-day, is still "in." The total available capacity from Plants A through E is 10,500 MW.

The Clearing Price: Where Supply Meets Demand

The auction "clears" at the price where the amount of offered capacity is just enough to meet PJM's required capacity.

In our example, when the price dropped and Plant F exited, there was still 10,500 MW of capacity available from the remaining plants (A, B, C, D, and E). Since this is the first point at which the available capacity meets or just exceeds the 10,000 MW requirement, the auction is over.

The clearing price is set by the offer of the last generator needed to meet the demand. In this case, that is Plant E, the natural gas peaker plant. Therefore, the clearing price for all successful participants is $150/MW-day.

The Results: Who Gets Paid and How Much

Here's the final outcome of our simplified auction:

• Cleared Generators: Plants A, B, C, D, and E.

• Uncleared Generators: Plants F and G did not clear because their offer prices were too high. They will not receive a capacity payment.

• The Payment: All the cleared generators (A through E) will receive the same clearing price of $150/MW-day for their committed capacity for the specified delivery year. This is a key feature of the auction – even though Plant A offered its capacity at a much lower price, it still receives the higher, market-clearing price.

This example demonstrates how the PJM capacity auction is designed to secure the necessary amount of power at the most competitive price, with the final clearing price being set by the last resource needed to ensure grid reliability.

Read More

US Power Markets

 

US Power Markets

PJM Service Area (inc. DC)

Read More

The Future of AI: Deployment, Disruption, and Dystopia?

The world of Artificial Intelligence (AI) is evolving at an unprecedented pace, even surprising industry insiders. This rapid acceleration is particularly evident in breakthroughs in next-generation AI and embodied AI, such as humanoid robotics, as well as the advanced development of AI agents.

The Future of AI: Deployment, Disruption, and Dystopia?

We can anticipate a broad deployment of AI applications in 2026 and 2027. We're already witnessing the initial phases of this integration, with some companies experiencing layoffs and the incorporation of AI into their business models. A recent example is Business Insider, which reportedly laid off 20% of its workforce. Interestingly, some early adopters are experiencing a sense of "regret," realizing that human involvement is still crucial for AI's success.

The 2020s are poised to be a period of massive innovation in AI. However, projections suggest the 2030s could lean towards a more dystopian future, potentially necessitating a fundamental shift in the social contract. Looking further ahead, the 2040s might bring about a period of abundance, positively impacting both work and leisure.

The Lag in AI Regulation

The current pace of AI regulation is significantly lagging behind technological advancements, and it appears this trend will continue. A sovereign arms race in AI is likely to take precedence over ethical considerations. Many major risks are currently being "auto-regulated" by individual companies. It's crucial to acknowledge that AI can exhibit behaviors as detrimental as humans, highlighting the urgent need to instill morality within AI systems.

Current Leaders in the AI Landscape

Google is viewed very positively, having successfully delivered usable AI to consumers. Companies like Apple and Intel are perceived as being behind in the AI race. Nvidia is considered to be performing well. The rumored merger between OpenAI and io is believed to be focused on acquiring an ambient AI-related device.

DeepSeek's Impact and the Future of AI Scaling

DeepSeek did not prove to be the disruption many anticipated. Its emergence demonstrated that simply scaling AI through more training and data is not the future of the technology. There are natural limits to the availability of information for AI training, and in the future, the marginal value of information and knowledge is expected to fall to zero. The availability of lower-cost AI will inevitably lead to increased demand and more diverse use cases. Nvidia, for instance, believes the market will absorb all the chips it can produce. The question of whether we have enough data centers remains an open one, reflecting the speculative nature of AI's future importance.

The Sovereign AI Race

The sovereign AI race is a tangible reality and will likely overshadow ethical concerns. China possesses approximately half of the world's AI researchers and can leverage its state control to its advantage. China is also reportedly leading in terms of power and compute capabilities. However, the United States still holds an edge with superior AI models and excellent access to capital for funding AI initiatives.

How AI Will Reshape Our World

AI will fundamentally change the world by driving the cost of knowledge to zero. AI businesses will likely adopt a tiered model to maintain profitability. Applications as we know them may fade into the background, with users relying on an AI assistant to "search" for them. Business costs, including labor and capital, are expected to decrease significantly.

Ultimately, AI is an amplifier of human talent. The key differentiator between businesses will remain human talent, as technology becomes increasingly accessible to everyone. While AI promises immense benefits, it will also introduce dystopian and ethical risks.

Bottom Line

• AI chip and energy demand are projected to remain high, as lower-cost solutions will only fuel increased demand.
• Labor disruption is likely to be very significant and occur in the near future.
• Productivity and profitability are expected to increase very soon.

Read More

OPINION: Supply Chain Woes

Supply chain risk was a child of COVID-19. With mass lockdown, industrial closures, limitations on transportation and shipping, supply chains were stretched. This was then exacerbated post COVID-19 as the world ramped back up to normal and supply chains unable to cope. This manifested in delivering delays and higher transportation costs along the entire value chain.

Now geopolitics are entering the mix of reasons for supply chains disruption. Protectionism and tariffs are increasing costs - project developers have been bulk ordering to try and get ahead of the curve - and reshoring are pushing up costs further. Restrictions on exports are adding to the woes. Think US CHIPS Act and Chinese restrictions on rare earths exports.

All this at a time when the world needs an affordable and functioning supply chain to deliver the energy transition, growth in power demand and AI compute.

This is what happens when supply chains were built for just-in-time to slim down costs and working capital. What the world needs now to rebuild is a just-in-case supply chain.

Read More

Real-time power consumption from two US hyperscale data centres

 

Real-time power consumption from two US hyperscale data centres (MW)

Read More

US Large Load Tariffs

 Large load datacenters are driving a large proportion of investment by utilities required for load augmentation. This introduces asymmetry with significant long-term investment to serve one specific group of customer over the current broad retail/commercial customer base.

This one customer group - hyperscalers - are also prone to rapid technological change. This poses the dilemma of how to ensure this group pays for the cost of serving them and shielding existing customers from the risk of future datacenter power demand being lower than expected.

Utilities are adopting several mechanisms to balance attracting new large customers while protecting existing customers (ratepayers) with "large load tariffs":

• Rates based on the marginal cost of serving the new customer
• Long-term contracts obliging payment of service regardless of whether power is required, to provide revenue certainty; an option to exit the contract for a fee could be a feature
• Minimum monthly demand and energy charges - i.e. take-or-pay, so that large customers contribute to grid costs even during low usage periods. Foe example, AEP Indiana uses a charge based on 80% of contracted or historical peak demand
• Collateral requirements

Examples of large load tariff protectionist features include:

State	Utility	Tariff Features	Purpose / Notes
New Mexico	Multiple Utilities	Special rates allowed by law if they recover incremental service costs	Supports economic development while protecting existing ratepayers
Indiana	AEP Indiana	Minimum demand charge (80% of contracted/historical peak)	Ensures cost recovery even during low usage periods
Kansas	Evergy	15-year contracts, collateral requirements, early termination fees	Provides revenue certainty and risk mitigation
Georgia	Georgia Power	Load forecast shows high growth; cautious tariff commitments due to project risk	Focus on data centers; many projects not yet committed
Texas	Oncor, CenterPoint	Tariffs include peak demand incentives and grid contribution requirements	Designed to manage grid stress and incentivize off-peak usage
Virginia	Dominion Energy	Tariffs include clean energy integration options and flexible load management	Aligns with state clean energy goals and large customer flexibility
Arizona	APS, SRP	Tariffs include time-of-use rates and infrastructure cost-sharing	Encourages load shifting and shared investment in grid upgrades
North Carolina	Duke Energy	Tariffs include customer commitment thresholds and performance guarantees	Ensures reliability and investment justification

Read More