pith. sign in

arxiv: 2606.00811 · v1 · pith:DZ3ERUFQnew · submitted 2026-05-30 · 💰 econ.EM · cs.AI

Certificates without Electrons? Theory and Evidence on Impacts from AI-Driven Power Demand

Dana Golden , Aruna Balasubramanian , Niranjan Balasubramanian This is my paper

Pith reviewed 2026-06-28 18:01 UTC · model grok-4.3

classification 💰 econ.EM cs.AI
keywords AI data centersrenewable energy certificateselectricity grid impactsdifference-in-differencestiming mismatchpower purchase agreementsfossil generation
0
0 comments X

The pith

Renewable certificates fail to prevent AI data centers from raising local fossil generation, prices, and outages due to timing mismatches.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that renewable certificates and power purchase agreements do not prevent AI data centers from increasing local fossil generation, electricity prices, and outages. A game-theoretic model identifies a timing wedge between real-time consumption and credited renewable output as the mechanism driving these effects even under full annual coverage. Using staggered large language model releases for causal identification via difference-in-differences, the analysis finds price increases up to 25 percent and 0.5 to 1 additional outages per year near data centers, with effects scaling by model size. On-site generation reverses some negative power-quality effects, and counterfactuals show that colocated storage and spatial reallocation reduce impacts while REC-only approaches do not.

Core claim

Even when renewable energy certificates cover 100 percent of annual data center consumption, the timing mismatch between AI-driven electricity use and credited renewable generation leads to higher fossil fuel generation, elevated wholesale prices, and increased outage frequency near data centers; behind-the-meter colocation mitigates these effects by absorbing demand spikes and encouraging renewable entry.

What carries the argument

The timing wedge between consumption and renewable crediting in the game-theoretic model of procurement choices and generator entry decisions.

Load-bearing premise

The staggered release of large language models provides exogenous variation in AI-driven electricity demand suitable for causal identification of local grid impacts via difference-in-differences.

What would settle it

No differential increase in fossil generation, wholesale prices, or outage frequency near data centers following large language model releases would falsify the central claim.

Figures

Figures reproduced from arXiv: 2606.00811 by Aruna Balasubramanian, Dana Golden, Niranjan Balasubramanian.

Figure 1
Figure 1. Figure 1: Projected data center growth over time. Source: DOE. around data centers owned by model developers? 3) How do these impacts translate into price effects in the wholesale and retail markets? Additional demand will likely increase prices, but understanding the magnitude of these price increases is important to determining the severity of the issue. 4) How will these impacts be different based on counterfactu… view at source ↗
Figure 2
Figure 2. Figure 2: Difference-in-differences estimation. Under the parallel trends assumption, the counterfactual (dashed) projects treatment group outcomes absent intervention. The ATT is identified as the divergence between observed and counterfactual outcomes post-treatment. Third, the switching power supplies and electronics essential to AI hardware generate harmonic distortion, introducing waveform distortions that spre… view at source ↗
Figure 6
Figure 6. Figure 6: Total harmonic distortion representation [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 3
Figure 3. Figure 3: Stacked DiD coefficients for power demand (IV specification). Event Time 0 corresponds to AI model publication date. The pre-treatment coefficient at 𝑡 = −2 demonstrates the training demand. Reference period is 𝑡 = −1 [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The map shows estimated AI-related wholesale prices increases in the PJM Interconnection in treated zones. First Author et al.: Preprint submitted to Elsevier Page 20 of 32 [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Effects of Geographic Concentration and On-site Generation on AI Impacts on the grid. First Author et al.: Preprint submitted to Elsevier Page 21 of 32 [PITH_FULL_IMAGE:figures/full_fig_p022_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Data centers by state. First Author et al.: Preprint submitted to Elsevier Page 22 of 32 [PITH_FULL_IMAGE:figures/full_fig_p023_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Map of zones for demand model [PITH_FULL_IMAGE:figures/full_fig_p024_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Map of Retail Electric Utilities for Whisker Labs Data. First Author et al.: Preprint submitted to Elsevier Page 23 of 32 [PITH_FULL_IMAGE:figures/full_fig_p024_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Map of PJM by Zones. First Author et al.: Preprint submitted to Elsevier Page 24 of 32 [PITH_FULL_IMAGE:figures/full_fig_p025_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Data sources and analysis pipeline. Data sources flow through variable construction into three estimation strategies addressing power quality impacts (Q1), fossil fuel demand (Q2), and counterfactual scaling scenarios (Q3) [PITH_FULL_IMAGE:figures/full_fig_p034_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: PJM Price Elasticity by Zone First Author et al.: Preprint submitted to Elsevier Page 33 of 32 [PITH_FULL_IMAGE:figures/full_fig_p034_12.png] view at source ↗
read the original abstract

Data centers now account for 4.4% of United States electricity demand, yet the grid-level effectiveness of the renewable energy certificates (RECs) and power purchase agreements (PPAs) hyperscalers use to claim carbon neutrality remains unclear. We develop a game-theoretic model in which a data center operator chooses among RECs, PPAs, and behind-the-meter colocation while generators make entry decisions under endogenous financing costs. The model identifies a timing wedge -- the mismatch between consumption and credited renewable generation -- as a central mechanism through which AI demand degrades reliability, raises prices, and increases emissions even when RECs cover 100% of annual consumption. Colocation with storage addresses this wedge directly and induces the greatest renewable entry by eliminating generator revenue risk. We test these predictions by exploiting the staggered release of large language models as a natural experiment, using difference-in-differences on a novel dataset linking AI activity to local grid outcomes. AI demand significantly increases fossil generation, wholesale prices (up to 25% in treated PJM zones), and outage frequency (0.5--1 additional outages per year) near data centers, with impacts scaling in model size. Data centers with on-site generation exhibit a sign reversal in power-quality effects, consistent with the model's prediction that behind-the-meter capacity absorbs demand spikes. Counterfactual analyses show that edge inference, spatial reallocation, and colocated storage each substantially mitigate grid impacts, while REC-only strategies do not. Together, our results demonstrate that the externalities of AI to the grid are tightly coupled to procurement design and the spatial organization of data center infrastructure.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper develops a game-theoretic model of data center operators choosing among RECs, PPAs, and behind-the-meter colocation, with generators making entry decisions under endogenous financing costs; the model highlights a timing wedge between consumption and credited renewable generation as the mechanism driving reliability, price, and emissions effects even under 100% REC coverage. It then implements a difference-in-differences design exploiting staggered large language model releases as a natural experiment on a novel dataset linking AI activity to local PJM grid outcomes, reporting that AI demand raises fossil generation, wholesale prices (up to 25% in treated zones), and outage frequency (0.5–1 additional per year), with effects scaling in model size and sign reversal for on-site generation; counterfactuals indicate mitigation from edge inference, spatial reallocation, and colocated storage but not from REC-only strategies.

Significance. If the identification strategy is valid, the results would provide evidence that AI-driven demand creates grid externalities tightly linked to procurement design and spatial organization, with implications for renewable entry incentives and reliability policy. The theoretical timing-wedge mechanism and the counterfactual analyses on mitigation strategies would be policy-relevant contributions to the literature on data-center electricity impacts.

major comments (2)
  1. [Empirical Strategy] The central DiD identification in the empirical section relies on staggered LLM releases delivering exogenous local demand shocks, but the manuscript provides no reported pre-trend diagnostics, tests for correlation between release timing and contemporaneous grid/economic shocks, or details on how global model releases are mapped to zone-level treatment intensity via the novel dataset; without these, the attribution of price, generation, and outage changes to AI demand cannot be assessed as causal.
  2. [Empirical Results] The claim that data centers with on-site generation exhibit a sign reversal in power-quality effects is load-bearing for the model's prediction on colocation, yet the manuscript supplies no robustness checks on how on-site generation is measured or coded, nor evidence against spillovers across PJM zones or anticipation effects around release dates.
minor comments (1)
  1. [Model] The abstract and model section would benefit from explicit notation for the timing wedge (e.g., defining the mismatch parameter) to improve readability for readers unfamiliar with the procurement instruments.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which identify key areas where additional empirical diagnostics can strengthen the paper. We address each point below and commit to revisions that incorporate the suggested checks.

read point-by-point responses

  1. Referee: [Empirical Strategy] The central DiD identification in the empirical section relies on staggered LLM releases delivering exogenous local demand shocks, but the manuscript provides no reported pre-trend diagnostics, tests for correlation between release timing and contemporaneous grid/economic shocks, or details on how global model releases are mapped to zone-level treatment intensity via the novel dataset; without these, the attribution of price, generation, and outage changes to AI demand cannot be assessed as causal.

    Authors: We agree these diagnostics are important for assessing causality. In the revision we will add event-study plots documenting pre-trends for all main outcomes, report pairwise correlations between LLM release dates and contemporaneous PJM/economic variables, and include an appendix detailing the mapping from global model releases to zone-level treatment intensity (using data-center locations, model parameter counts, and timing). These additions will be placed in the main empirical section and appendix. revision: yes

  2. Referee: [Empirical Results] The claim that data centers with on-site generation exhibit a sign reversal in power-quality effects is load-bearing for the model's prediction on colocation, yet the manuscript supplies no robustness checks on how on-site generation is measured or coded, nor evidence against spillovers across PJM zones or anticipation effects around release dates.

    Authors: We will add the requested robustness material. The revision will report results under alternative codings of on-site generation (different capacity thresholds and data sources), include neighboring-zone placebo tests to assess spillovers, and add leads of the treatment indicators to examine anticipation around release dates. These checks will appear alongside the main sign-reversal results. revision: yes

Circularity Check

0 steps flagged

No significant circularity; theoretical model and DiD estimates rely on external timing variation.

full rationale

The paper first presents a game-theoretic model deriving predictions about a timing wedge, reliability impacts, and mitigation via colocation/storage. It then tests those predictions empirically via difference-in-differences that exploit the staggered release dates of large language models as exogenous variation in AI-driven demand. This identifying variation is external to the model parameters and data-center outcomes, with no equations or claims showing that empirical estimates reduce to fitted inputs, self-definitional constructs, or load-bearing self-citations. The derivation chain therefore contains independent empirical content against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based solely on the abstract; the model rests on standard domain assumptions about operator and generator behavior with no new entities introduced.

axioms (2)
  • domain assumption Data center operators choose among RECs, PPAs, and behind-the-meter colocation while generators make entry decisions under endogenous financing costs.
    Core setup of the game-theoretic model described in the abstract.
  • domain assumption A timing wedge between consumption and credited renewable generation drives reliability, price, and emission effects.
    Identified as the central mechanism in the abstract.

pith-pipeline@v0.9.1-grok · 5829 in / 1315 out tokens · 32581 ms · 2026-06-28T18:01:31.563496+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

179 extracted references · 16 canonical work pages · 2 internal anchors

  1. [1]

    Communications of the ACM , volume=

    Green ai , author=. Communications of the ACM , volume=. 2020 , publisher=

    2020

  2. [2]

    Energy and

    Strubell, Emma and Ganesh, Ananya and McCallum, Andrew , year = 2019, booktitle =. Energy and

    2019

  3. [3]

    Available at SSRN 4708811 , year =

    Backstrom, Jesse D and Gillenwater, Michael and Brander, Matthew and Inman, Charlie , title =. Available at SSRN 4708811 , year =

  4. [4]

    Available at SSRN 4591413 , year =

    Backstrom, Jesse D and Gillenwater, Michael and Inman, Charlie and Brander, Matthew , title =. Available at SSRN 4591413 , year =

  5. [5]

    Estimating the environmental impact of Generative-AI services using an LCA-based methodology , journal =

    Berthelot, Adrien and Caron, Eddy and Jay, Mathilde and Lef\`. Estimating the environmental impact of Generative-AI services using an LCA-based methodology , journal =. 2024 , type =

    2024

  6. [6]

    2024 , type =

    EPRI , title =. 2024 , type =

    2024

  7. [7]

    2024 , type =

    IEA , title =. 2024 , type =

    2024

  8. [8]

    Proceedings of the 15th ACM International Conference on Future and Sustainable Energy Systems , year =

    Maji, Diptyaroop and Bashir, Noman and Irwin, David and Shenoy, Prashant and Sitaraman, Ramesh K , title =. Proceedings of the 15th ACM International Conference on Future and Sustainable Energy Systems , year =

  9. [9]

    Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 3 , year =

    Patel, Pratyush and Choukse, Esha and Zhang, Chaojie and Goiri, Inigo and Warrier, Brijesh and Mahalingam, Nithish and Bianchini, Ricardo , title =. Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 3 , year =

  10. [10]

    Computer , volume =

    Patterson, David and Gonzalez, Joseph and Hoelzle, Urs and Le, Quoc and Liang, Chen and Munguia, Lluis-Miquel and Rothchild, Daniel and So, David R and Texier, Maud and Dean, Jeff , title =. Computer , volume =. 2022 , type =

    2022

  11. [11]

    Carbon Emissions and Large Neural Network Training

    Patterson, David and Gonzalez, Joseph and Le, Quoc and Liang, Chen and Munguia, Lluis-Miquel and Rothchild, Daniel and So, David and Texier, Maud and Dean, Jeff , title =. arXiv preprint arXiv:2104.10350 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  12. [12]

    2022 IEEE High Performance Extreme Computing Conference (HPEC) , publisher =

    Shankar, Sadasivan and Reuther, Albert , title =. 2022 IEEE High Performance Extreme Computing Conference (HPEC) , publisher =. 2022 , pages =

    2022

  13. [13]

    2024 , type =

    Shehabi, Arman and Hubbard, Alex and Newkirk, Alex and Lei, Nuoa and Siddik, Md Abu Bakkar and Holecek, Billie and Koomey, Jonathan and Masanet, Eric and Sartor, Dale , title =. 2024 , type =

    2024

  14. [14]

    Proceedings of Machine Learning and Systems , volume =

    Wu, Carole-Jean and Raghavendra, Ramya and Gupta, Udit and Acun, Bilge and Ardalani, Newsha and Maeng, Kiwan and Chang, Gloria and Aga, Fiona and Huang, Jinshi and Bai, Charles , title =. Proceedings of Machine Learning and Systems , volume =. 2022 , type =

    2022

  15. [15]

    Energy Economics , volume=

    Energy transition and non-energy firms' financial performance: Do markets value capability-based energy transition strategies? , author=. Energy Economics , volume=. 2024 , publisher=

    2024

  16. [16]

    American Economic Review , volume=

    Design-based research in empirical microeconomics , author=. American Economic Review , volume=. 2022 , publisher=

    2022

  17. [17]

    ILR Review , volume=

    Comment on David Neumark and William Wascher,``Employment effects of minimum and subminimum wages: Panel data on state minimum wage laws'' , author=. ILR Review , volume=. 1994 , publisher=

    1994

  18. [18]

    The Thirteenth International Conference on Learning Representations (ICLR) Spotlight , year =

    Holistically Evaluating the Environmental Impact of Creating Language Models , author =. The Thirteenth International Conference on Learning Representations (ICLR) Spotlight , year =. 2503.05804 , archiveprefix =

    work page arXiv

  19. [19]

    2025 , eprint =

    The Cost of Dynamic Reasoning: Demystifying AI Agents and Test-Time Scaling from an AI Infrastructure Perspective , author =. 2025 , eprint =

    2025

  20. [20]

    2025 , eprint =

    Measuring the Environmental Impact of Delivering AI at Google Scale , author =. 2025 , eprint =

    2025

  21. [21]

    Bloomberg , year =

    Leonardo Nicoletti and Naureen Malik and Andre Tartar , title =. Bloomberg , year =

  22. [22]

    The Economics of New Goods , volume=

    Valuation of new goods under perfect and imperfect competition , author=. The Economics of New Goods , volume=. 1997 , publisher=

    1997

  23. [23]

    American Economic Review , volume=

    What do emissions markets deliver and to whom? Evidence from Southern California's NOx trading program , author=. American Economic Review , volume=

  24. [24]

    American Economic Review , volume=

    Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania , author=. American Economic Review , volume=

  25. [25]

    American Economic Review , volume=

    Environmental regulations, air and water pollution, and infant mortality in India , author=. American Economic Review , volume=

  26. [26]

    Energies , volume=

    Sustainable energy data centres: A holistic conceptual framework for design and operations , author=. Energies , volume=. 2023 , publisher=

    2023

  27. [27]

    arXiv preprint arXiv:2411.09786 , year=

    Environmental burden of United States data centers in the artificial intelligence era , author=. arXiv preprint arXiv:2411.09786 , year=

    work page arXiv

  28. [28]

    Computational Intelligence for Green Cloud Computing and Digital Waste Management , pages=

    Impact of data centers on power consumption, climate change, and sustainability , author=. Computational Intelligence for Green Cloud Computing and Digital Waste Management , pages=. 2024 , publisher=

    2024

  29. [29]

    2025 , note =

    Aterio , title =. 2025 , note =

    2025

  30. [30]

    Electric Grid , institution =

    Analysis of Total Harmonic Distortion on the U.S. Electric Grid , institution =. 2024 , month =

    2024

  31. [31]

    2022 , note =

    United States Environmental Protection Agency , title =. 2022 , note =

    2022

  32. [32]

    2009 , journal =

    National Association of Regulatory Utility Commissioners , title =. 2009 , journal =

    2009

  33. [33]

    2015 , institution=

    Natural gas prices and coal displacement: Evidence from electricity markets , author=. 2015 , institution=

    2015

  34. [34]

    American Economic Review , volume=

    Imperfect markets versus imperfect regulation in US electricity generation , author=. American Economic Review , volume=. 2022 , publisher=

    2022

  35. [35]

    2024 , note =

    EIA , title =. 2024 , note =

    2024

  36. [36]

    Environmental and Resource Economics , volume=

    The endogeneity problem in applied fisheries econometrics: A critical review , author=. Environmental and Resource Economics , volume=. 2015 , publisher=

    2015

  37. [37]

    2025 , month = may, url =

    2025 Summer Reliability Assessment , institution =. 2025 , month = may, url =

    2025

  38. [38]

    2024 , month = jul, type =

    Dan Thompson , title =. 2024 , month = jul, type =

    2024

  39. [39]

    2014 , address =

    IEEE Recommended Practice and Requirements for Harmonic Control in Electric Power Systems , number =. 2014 , address =

    2014

  40. [40]

    2024 , month =

    Theo Laughner and Stan Heckman and Angel Manzur and Chris Sloop , title =. 2024 , month =

    2024

  41. [41]

    2023 , note =

    Table 4.2.A: Existing Net Summer Capacity by Energy Source and Producer Type, 2013 through 2023 (Megawatts) , howpublished =. 2023 , note =

    2013

  42. [42]

    2024 United States Data Center Energy Usage Report , institution =

    Shehabi, Arman and Smith, Sarah and Sartor, Dale and Brown, Richard and Herrlin, Magnus and Koomey, Jonathan and Masanet, Eric and Horner, Nathaniel and Azevedo, In. 2024 United States Data Center Energy Usage Report , institution =. 2024 , month =

    2024

  43. [43]

    2024 , url =

    Powering Intelligence: Analyzing Artificial Intelligence and Data Center Energy Consumption , institution =. 2024 , url =

    2024

  44. [44]

    2025 , month =

    Goggin, Michael and Gramlich, Rob , title =. 2025 , month =

    2025

  45. [45]

    Bloomberg , year =

    Nicoletti, Leonardo and Malik, Naureen and Tartar, Andre , title =. Bloomberg , year =

  46. [46]

    Electricity Grid Impacts of Rising Demand from Data Centers and Cryptocurrency Mining Operations , institution =

    Luby, Ian and Sergi, Brian and Azevedo, In. Electricity Grid Impacts of Rising Demand from Data Centers and Cryptocurrency Mining Operations , institution =. 2025 , url =

    2025

  47. [47]

    2025 , month =

    Jacobs, Mike , title =. 2025 , month =

    2025

  48. [48]

    and Harvey, Scott M

    Hogan, William W. and Harvey, Scott M. , title =. The Electricity Journal , year =

  49. [49]

    2022 , url =

    Price Formation and Grid Operation Impacts from Variable Renewable Energy , institution =. 2022 , url =

    2022

  50. [50]

    Joule , volume =

    de Vries, Alex , title =. Joule , volume =. 2023 , doi =

    2023

  51. [51]

    and Bilinski, Alyssa and Poe, John , title =

    Roth, Jonathan and Sant'Anna, Pedro H.C. and Bilinski, Alyssa and Poe, John , title =. Journal of Econometrics , volume =. 2023 , doi =

    2023

  52. [52]

    Journal of econometrics , volume=

    Difference-in-differences with variation in treatment timing , author=. Journal of econometrics , volume=. 2021 , publisher=

    2021

  53. [53]

    Treatment Allocation under Uncertain Costs

    Treatment allocation under uncertain costs , author=. arXiv preprint arXiv:2103.11066 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  54. [54]

    Journal of econometrics , volume=

    Difference-in-differences with multiple time periods , author=. Journal of econometrics , volume=. 2021 , publisher=

    2021

  55. [55]

    American economic review , volume=

    Two-way fixed effects estimators with heterogeneous treatment effects , author=. American economic review , volume=. 2020 , publisher=

    2020

  56. [56]

    Review of Economic Studies , volume=

    Revisiting event-study designs: robust and efficient estimation , author=. Review of Economic Studies , volume=. 2024 , publisher=

    2024

  57. [57]

    The Quarterly Journal of Economics , volume=

    The effect of minimum wages on low-wage jobs , author=. The Quarterly Journal of Economics , volume=. 2019 , publisher=

    2019

  58. [58]

    American Economic Journal: Economic Policy , volume=

    Who is screened out? Application costs and the targeting of disability programs , author=. American Economic Journal: Economic Policy , volume=. 2019 , publisher=

    2019

  59. [59]

    American Economic Review , volume=

    Measuring duopoly power in the British electricity spot market , author=. American Economic Review , volume=. 1999 , publisher=

    1999

  60. [60]

    American Economic Review , volume=

    Measuring market inefficiencies in California's restructured wholesale electricity market , author=. American Economic Review , volume=. 2002 , publisher=

    2002

  61. [61]

    2023 , month = sep, howpublished =

    2023

  62. [62]

    2023 , month = apr, howpublished =

    2023

  63. [63]

    arXiv preprint arXiv:2211.02001 , year =

    Luccioni, Alexandra Sasha and Viguier, Sylvain and Ligozat, Anne-Laure , title =. arXiv preprint arXiv:2211.02001 , year =

    work page arXiv

  64. [64]

    2023 , month = may, howpublished =

    2023

  65. [65]

    2023 , month = dec, howpublished =

    Kurian, Thomas , title =. 2023 , month = dec, howpublished =

    2023

  66. [66]

    2025 , month = nov, howpublished =

    2025

  67. [67]

    2021 , howpublished =

    2021

  68. [68]

    2020 , url =

    OpenAI API , author =. 2020 , url =

    2020

  69. [69]

    2021 , url =

    API No Waitlist , author =. 2021 , url =

    2021

  70. [70]

    2021 , url =

    Introducing OpenAI Codex , author =. 2021 , url =

    2021

  71. [71]

    2022 , url =

    DALL-E API Now Available in Public Beta , author =. 2022 , url =

    2022

  72. [72]

    2022 , url =

    ChatGPT , author =. 2022 , url =

    2022

  73. [73]

    2023 , url =

    Introducing ChatGPT and Whisper APIs , author =. 2023 , url =

    2023

  74. [74]

    2023 , url =

    GPT-4 , author =. 2023 , url =

    2023

  75. [75]

    2023 , url =

    GPT-4 API General Availability , author =. 2023 , url =

    2023

  76. [76]

    2023 , url =

    GPT-4 System Card , author =. 2023 , url =

    2023

  77. [77]

    2023 , url =

    New Models and Developer Products Announced at DevDay , author =. 2023 , url =

    2023

  78. [78]

    2024 , url =

    New Embedding Models and API Updates , author =. 2024 , url =

    2024

  79. [79]

    2024 , url =

    Hello GPT-4o , author =. 2024 , url =

    2024

  80. [80]

    2024 , url =

    GPT-4o mini: Advancing Cost-Efficient Intelligence , author =. 2024 , url =

    2024

Showing first 80 references.