Mike Grunow

Energy Abundance

[Mike Grunow @MichaelGrunow]

@KalshiFB Drew Bledsoe never injured. No Brady.

[Mike Grunow @MichaelGrunow]

@HawkEmDownChris Chad Pennington

[Mike Grunow @MichaelGrunow]

@achester99 @JonMetler Phil Simms is 10.

[Mike Grunow @MichaelGrunow]

@RespectableCon I look for wealthy counties and hope to see non-red. For example, Fairfield county CT. Consisting of Greenwich, Darien, New Canaan. Would prefer they be Green, not non-red is a start in a sea of red in the Northeast.

[Mike Grunow @MichaelGrunow]

@ShanuMathew93 What’s the effective all in toll in $/kw-mo?

[Mike Grunow @MichaelGrunow]

@AccentsSouthern WMFP

[Jamison Daniel @AntiquarianMuse]

A fully masked ICE agent pepper sprayed a woman in New York yesterday merely for videotaping
He approached her. She was standing still. That guy needs jail time.

If this goes viral, the city will seek charges.

[Mike Grunow @MichaelGrunow]

@kylefromvista @mikegolicjr Yup

[Mike Grunow @MichaelGrunow]

@njrealtornic We doing this again?!?!?

[Mike Grunow @MichaelGrunow]

@NYNJHarper I enjoyed winning the Philly series. We were one hit away all night from taking 2 out of 3 vs dodgers but Soto was pulled. The team is competing post ASG

[Mike Grunow @MichaelGrunow]

@mrmikeMTL Solar panels, batteries

[Mike Grunow @MichaelGrunow]

@Willpa11 I fully expect 2 season crippling injuries

[Duncan S. Campbell @duncancampbell]

Here’s three ways Scale tests capacity reliability on every data center power project, from simple to complex:

1) N-k Analysis

An N-k analysis is used to test whether the system has enough installed capacity to continue serving the required load after losing a selected number of generating units or other critical components. We start with the full system capacity, remove (k) components (typically using a defined or conservative outage combination) and then compare the remaining available capacity against the load requirement. By varying (k), we can see how much redundancy the system actually has and identify the point at which an additional outage would create a capacity shortfall. And if you attach probability of failure to your generators, you can use this approach to create a coarse stochastic view of capacity availability at varying load levels.

2) Capacity Outage Probability Table (COPT) Analysis

Rather than assume a fixed quantity of concurrent forced outages, as in the N-k approach, a COPT analysis assigns an annual availability (typically based on a manufacturer guarantee) and a deliverable capacity to each generator type being used in the project. Next, it calculates the probability of each possible combination of generator outages across the full fleet, including systems with multiple generator types, unit sizes, and availability assumptions. For each outage state, the remaining available capacity is compared against the required load, allowing us to develop a finer-grained view of the percentage of load served and the expected annual availability delivered by a given generation solution. This approach is typically used to quickly determine what peak load can be served at an availability target using a given generation solution.

3) Time-Series Power Balance & Maintenance-Informed RAM Analysis

This approach begins with a time-series model of the generation system operating against representative AI workload shapes, while explicitly scheduling and performing every required maintenance activity, including when it must occur and how long each piece of equipment must remain unavailable. Establishing this deterministic baseline is important because many generator maintenance activities are triggered by operating-hour thresholds, meaning that maintenance requirements can become correlated across units that have accumulated similar operating hours and may cluster within the same periods. The model therefore creates a deterministic view of the operating state of every major component over a 20-year study period, including all planned outages, and shows the spare capacity available relative to load at every moment. Next, a stochastic RAM analysis is layered onto that baseline using a model of the complete electrical topology and component-specific mean-time-between-failure and mean-time-to-repair distributions. The model runs millions of Monte Carlo simulations, testing whether sufficient power can still reach the load whenever one or more components unexpectedly change state, while accounting for battery storage’s fast response, available headroom and ramp rates from online generators, and the startup time of offline generators. The result is a high-resolution probability distribution of system performance that allows the user to evaluate expected availability and load served at the confidence level most appropriate for the project and guarantees being made, such as P50, P75, P90, or P99.

[Mike Grunow @MichaelGrunow]

@emmitonair Antonio Cromartie

[Mike Grunow @MichaelGrunow]

@ShanuMathew93 Row 2 won't happen, the site will be expanded to 2 GWs.

[Mike Grunow @MichaelGrunow]

@ZachLowe_NBA We still watching zeke?

[Mike Grunow @MichaelGrunow]

@SeanCasten Until tomorrow

[Mike Grunow @MichaelGrunow]

@RichardMeyerDC Looks like our founding fathers were 100% renewable!

[Mike Grunow @MichaelGrunow]

@Dan_Canoeth @RichardMeyerDC That’s right. Amd most grids aren’t able to handle the ramps easily.

[Mike Grunow @MichaelGrunow]

@ShanuMathew93 @MegaWattXinfo The DC needs a BESS if grid tied or islanded.