How virtual power plants could provide energy for data centers

Would you take a payment to ramp down your electricity use? Would it change anything if you were doing so to help power a local data center?

Google just signed a new deal to help pay for a virtual power plant (VPP) in the largest power grid in the US. The agreement is with Voltus, a leading VPP and distributed energy resources platform.

Voltus will set up the virtual power plant, grouping together devices like electric vehicles and smart thermostats. It’ll pay customers to participate, and the company will dial back power or use the stored energy during times when the grid is stressed. Google will foot the bill for setting it up, and the extra capacity generated by the project will help run its data centers in the region.

This is one of the most concrete examples so far of a tech giant using a VPP to help meet energy demand for data centers. But there are still some lingering questions about just how far this sort of program can go, and what the limits are.

Last year, it felt as if everyone was talking about data center flexibility. A high-profile study from Duke University found that if data centers agreed to decrease their energy demand for roughly 40 hours per year, a whole bunch of them (about 100 gigawatts’ worth) could come online without making new power plants or transmission equipment necessary.

The underlying reason is that our power grid is designed not for our average energy use, but for the absolute maximum: the brutally hot July evening when everyone is blasting their air conditioners, watching Love Island, and microwaving popcorn. If a data center is willing to refrain from pulling so much power during those high-stress times, the grid can happily support it the rest of the year.

One lingering question here is about incentives: How would you get data centers to agree to this? After all, they might not have a very flexible load, especially now that AI use is more widespread—training a model can easily be delayed or shifted, but customer demand is more immediate. Giving up computing capacity could mean losing revenue.

Regulation is one approach that could work here. One proposal in the US would allow new data centers to come online years sooner if they agree to lower demand when the grid is nearing its max.  And a new Texas law requires large users to switch to backup power or curtail their demand in emergency situations.

Another approach is for data center operators to pay for other people to be flexible.

Voltus announced a new program in September that allows data centers to finance flexibility on their local grid. The company calls it “Bring your own capacity.” Google is now the first named customer taking advantage of this program.

In the new agreement, Voltus will pay people who agree to participate in the virtual power plant. The plant will be part of PJM, the grid that covers much of the US East Coast. The company says it will be able to aggregate up to 100 megawatts of distributed energy resources each year. The plant should be operational in 2027, according to Voltus.

This isn’t Google’s first foray into flexibility; the company has agreements with utilities across the US to limit or shift its own energy demand, which can help free up grid capacity. As the company pointed out in a blog post earlier this year, though, there are limits on how flexible a data center can be, and not every facility will be able to ramp down its power demand.

“There is no one solution for expanding grid capacity and we’re continuing to explore all options, including the many avenues for load flexibility,” said Michael Terrell, Google’s global head of advanced energy, in an emailed statement in response to written questions.

Once again, I’m wondering about incentives here. These companies are asking homes and businesses to be flexible. Will they agree?

A recent study in California looked at local people’s willingness to participate in managed electric-vehicle charging. Essentially, the program pays people to give up control of when they charge their EVs. This is another way to help smooth out electricity demand and ease the burden on the grid.

The problem? Not many people signed up. With no economic incentive, only 1% of EV owners enrolled in managed charging. At $40 per month (about 15% of their power bill), only 4.6% did.

This is a different situation and a different region from the one in which Google is working with Voltus. (It’s worth noting that the companies aren’t sharing how much they plan to pay the participants, which will obviously be a big determinant in participation for this kind of project.) 

But this study shows that even with money on the table, people may not always jump at the chance to cede control of their electricity demand. And it certainly feels relevant that about 70% of Americans oppose AI data centers in their area, according to recent Gallup polling

Being flexible sounds like a great idea in theory, and these financed VPPs could provide an immediate route to meeting energy demand. But as we move from idea to implementation, it’ll be interesting to see whether trial runs work as intended.  

This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here

STAT+: What the pope’s encyclical on AI means for Catholic hospitals, and all of health care

You’re reading the web edition of STAT’s AI Prognosis newsletter, our subscriber-exclusive guide to artificial intelligence in health care and medicine. Sign up to get it delivered in your inbox every Wednesday. 

The sequel to last week’s graduation speeches that garnered boos for AI: Comedian Ronny Chieng addressed Harvard’s graduating class and told them their mission was to “destroy AI,” which prompted a roar of approval.

He wasn’t referring to uses of AI in medicine and physics, he specified, but rather offloading writing and creating to AI. The journey of making and learning and figuring out is “the point of all of this,” he said. “Why would I want AI to take that away from me?”

Continue to STAT+ to read the full story…

STAT+: Legend Biotech emerged as a rare market winner

Want to stay on top of the science and politics driving biotech today? Sign up to get our biotech newsletter in your inbox.

Good morning. Yesterday was a brutal day for biotech stocks, but one company withstood the broad selloff. We also look at investor excitement for longevity startups and yet another licensing deal made by Eli Lilly.

Legend Biotech emerged as a rare market winner

Shares of Legend Biotech soared over 40% yesterday after the company disclosed early data on its in vivo CAR-T therapy that showed promise in Non-Hodgkin’s lymphoma.

Continue to STAT+ to read the full story…

STAT+: Pharmalittle: We’re reading about GLP-1 drugs and knees, FDA cell and gene therapy guidance, and more

Good morning, everyone, and welcome to the middle of the week. Congratulations on making it this far, and remember there are only a few more days until the weekend arrives. So keep plugging away. After all, what are the alternatives? While you ponder the possibilities, we invite you to join us for a delightful cup of stimulation. Our choice today is maple cinnamon French toast. Meanwhile, here is the latest menu of tidbits to help you on your way. Have a wonderful day and please do stay in touch. …

Cigna will stop covering GLP-1 weight loss drugs including Novo Nordisk’s Wegovy and Eli Lilly’s Zepbound in its ​employee health plan effective July 1, Reuters reports. In a document ​circulated to employees on June 1, Cigna suggested those currently using the drug can choose to pay with cash through manufacturer sites or TrumpRx. The cash-pay purchases will not apply toward a deductible or the amount of ​spending required before enrollees can use their health coverage. The price of weight loss ​drugs has been falling in 2026 with the launch of Novo’s Wegovy pill and Lilly’s oral Foundayo, with ⁠prices that start at $149 per month for the lowest dose. Americans have been increasingly pushed to the cash-pay market and, at the same time, employers have been cutting back on their coverage of the drugs.

Taking weight loss drugs for at least three years could prevent thousands of knee replacements a year, The Guardian writes, citing new research. The study, published in Regional Anesthesia & Pain Medicine, found that taking GLP-1 medications for one year was associated with a 1.4% reduced risk of knee replacement surgery at the three-year follow-up point and a 2.8% lower risk after eight years. But the greatest reduction in risk was with newer weight loss drugs and longer treatment. Taking semaglutide or tirzepatide for three years was associated with a nearly 5% lower chance of needing knee replacement at the eight-year follow-up assessment.

Continue to STAT+ to read the full story…

Transcriptional signatures of the cortical morphometric similarity network gradient in left temporal lobe epilepsy with different seizure symptoms

BackgroundTemporal lobe epilepsy (TLE) manifests with diverse seizure symptoms, including focal to bilateral tonic—clonic seizures (FBTCS), linked to widespread brain network disruptions. The role of cortical morphometric similarity (MS) network gradients and their relationship with gene expression in TLE remains unclear.MethodsWe studied MS network gradient abnormalities through group comparisons among 87 left TLE patients (48 FBTCS−, 39 FBTCS+) and 63 healthy controls (HC). In addition, partial least squares (PLS) regression analysis was performed to investigate the association between gradient changes and whole-brain gene expression in left FBTCS+ TLE patients.ResultsFBTCS+ patients showed significant reductions in the principal MS network gradient within default mode network (DMN) regions compared to healthy controls, while FBTCS− patients exhibited no such abnormalities. Gradient alterations in FBTCS+ were linked to whole-brain expression of genes involved in neurobiological pathways, cell types, and cortical layers.ConclusionFBTCS+ TLE is associated with distinct MS network gradient alterations, which may reflect underlying molecular mechanisms contributing to structural changes linked to severe seizure symptoms.

Characteristic MRI pattern in LMNB1-related autosomal dominant leukodystrophy: a case report

IntroductionAdult-onset autosomal dominant leukodystrophy (ADLD) is an ultra-rare inherited white matter disorder caused by variants in the LMNB1 gene. Here, we report a case of ADLD and characterize its typical magnetic resonance imaging (MRI) features, with the aim of facilitating its clinical recognition and differential diagnosis.Case descriptionThe patient was a 55-year-old male who had experienced incomplete voiding, dysuria, and constipation for 10 years. One year prior to presentation, he developed lower limb weakness and unsteady gait, which progressively worsened over time. Brain MRI revealed extensive white matter abnormalities, including a symmetric hyperintensity pattern in the brainstem corticospinal tract and bilateral middle cerebellar peduncles on T2-weighted/FLAIR images, which resembled the facial profile of a Beagle dog. Subsequent genetic testing identified a pathogenic duplication of the LMNB1 gene, a typical variant associated with ADLD.ConclusionWe report a case of ADLD caused by LMNB1 duplication with a typical clinical course and characteristic MRI features. Its characteristic MRI features, including the “Beagle sign” as an illustrative imaging analogy in the brainstem, may facilitate the clinical recognition and differential diagnosis of this disorder.

A novel early screening approach for MCI due to AD based on a “maze” hand-interaction kinetic paradigm

BackgroundMild cognitive impairment due to Alzheimer’s disease (MCI due to AD) is a crucial stage for the early identification of Alzheimer’s disease (AD), and timely detection at this stage may provide opportunities for earlier intervention and potentially delay disease progression.MethodsThis study proposed a digital early-screening method based on a touchscreen maze hand-interaction kinetic paradigm, which integrates digital biomarkers from the visuospatial/executive and episodic memory domains to support the screening of MCI due to AD. A customized maze task was administered to 40 patients with clinically diagnosed MCI due to AD and 40 healthy controls (HC). Behavioral data were collected, and two categories of digital biomarkers were extracted: (1) visuospatial/executive digital biomarkers, such as task completion time (VSETT) and average movement speed (VSES); and (2) episodic memory digital biomarkers, such as episodic memory total time (EMTT) and number of correct choices (EMCC). Significant digital biomarkers were identified through between-group comparisons, and their combined classification performance was evaluated using binary logistic regression and receiver operating characteristic (ROC) analysis.ResultsThe integrated digital biomarker model showed promising apparent discriminative performance in the full cohort, with an AUC of 0.899 (95% CI: 0.831–0.967). To reduce potential optimism associated with biomarker selection, model development, and model evaluation within the same dataset, internal validation was performed using full-pipeline repeated stratified five-fold cross-validation with all 16 candidate digital biomarkers entered into the validation procedure and biomarker selection repeated within each training fold. The internally validated model retained good discriminative performance, with a mean cross-validated AUC of 0.842, an empirical 95% interval of 0.779–0.878, an accuracy of 0.783, a sensitivity of 0.772, and a specificity of 0.795.ConclusionThese findings suggest that the proposed touchscreen maze-based digital assessment method may provide a promising and objective approach to supporting the early screening of MCI due to AD.

Unraveling the role of asymmetric excitatory and inhibitory synaptic inputs to retinal ganglion cell direction selectivity

IntroductionUnderstanding the mechanisms underlying direction selectivity in retinal ganglion cells (RGCs) is crucial in visual neuroscience. Retinal direction selectivity is critical for gaze stabilization through optokinetic and vestibulo-ocular reflexes, and its loss impairs the ability to stabilize gaze and track moving objects, potentially impacting behaviors that rely on accurate motion detection. The prevailing hypothesis proposes that the interplay between excitatory and inhibitory inputs is pivotal for the emergence of direction selectivity in RGCs.MethodsTo dissect the contributions of these inputs, we employed dynamic-clamp recordings utilizing computationally modeled, synthetic excitatory and inhibitory conductances of varying amplitudes and time onsets in mouse RGCs.ResultsThe use of combinations of excitatory and inhibitory conductances with altered amplitude or timing in configurations not found in natural physiological conditions, allowed us to evaluate the specific contribution and impact of these modified components on direction selectivity. We found that asymmetries in both excitatory and inhibitory inputs are critical for the emergence of sharp directional tuning.DiscussionOur findings contribute to advance our understanding of the cellular and synaptic mechanisms that underlie retinal direction selectivity.

Rebuilding spinal circuit function after spinal cord injury through a patient-specific interneuron precision model

Spinal interneurons constitute the physiological core of spinal circuitry, integrating excitatory and inhibitory inputs to generate the rhythmic patterns that drive locomotor, postural, and autonomic control. Their developmental logic, molecular diversity, and adaptive plasticity make them central determinants of functional recovery after spinal cord injury (SCI). Yet most regenerative strategies continue to emphasize cellular replacement rather than the restoration of the physiological integrity of spinal networks. In this article, we reframe spinal repair as the restoration of interneuron-mediated circuit organization rather than cellular replacement alone. We synthesize current insights into how embryonic patterning programs defined by Sonic Hedgehog (SHH), Wnt, and bone morphogenetic protein (BMP) gradients, refined by Notch and retinoic acid signaling, and consolidated by axon guidance cues, establish interneuron diversity, connectivity, and network symmetry that together encode the logic of motor coordination. SCI disrupts this developmental logic, fragmenting excitatory and inhibitory balance and desynchronizing rhythmic modules, while residual circuits retain latent capacity for resynchronization through plasticity and neuromodulation. Building upon this developmental and physiological continuum, we propose the Patient-Specific Interneuron Precision Model (PIPM), a feedback-informed conceptual framework that links patient-specific biological states, including progenitor competence, morphogen sensitivity, metabolic tone, inflammatory burden, and lesion-specific circuit preservation, to circuit-level function and recovery potential. Frameworks such as the PIPM may help integrate molecular, physiological, and clinical dimensions of recovery, providing a path toward more personalized strategies for treating SCI through restoration of interneuron-mediated network organization.

Prediction model for postoperative delirium risk in elderly hypertensive patients: machine learning-based development and validation

BackgroundPostoperative delirium (POD) is a severe complication in elderly hypertensive patients, associated with poor long-term outcomes. Existing models often rely on intraoperative data, limiting preoperative risk stratification. This study aimed to develop a non-invasive machine learning model to predict POD and investigate its preoperative markers’ impact on three-year mortality.MethodsPreoperative variables were selected using LASSO regression from a cohort of 1,782 patients. Ten machine learning models were trained and validated (7:3 ratio). Model performance was evaluated via AUC-ROC and decision curve analysis (DCA). The optimal model was interpreted using SHAP values. Long-term prognosis within the POD cohort was assessed using Kaplan-Meier curves and multivariable Cox proportional hazards regression.ResultsThe POD incidence was 10.9%. The Gradient Boosting Machine (GBM) demonstrated optimal performance (AUC = 0.868, 95% CI: 0.819–0.917). SHAP analysis identified MMSE score as the most influential predictor, followed by HADS score, age, CFS, frailty, and PSQI score. Multivariable Cox analysis revealed that lower MMSE, alongside elevated HADS, CFS, frailty, and PSQI scores—but not chronological age—were independent predictors of increased three-year mortality in POD patients (all P < 0.05).ConclusionWe developed a robust machine learning tool for individualized POD prediction. Cognitive impairment, psychological distress, frailty, and poor sleep quality serve as critical dual-prognostic markers for both acute POD occurrence and long-term survival. These findings underscore the necessity of routine multidimensional preoperative assessment to facilitate personalized interventions for vulnerable hypertensive populations.