Universal photonic artificial intelligence acceleration (original) (raw)

Nature volume 640, pages 368–374 (2025)Cite this article

Subjects

Abstract

Over the past decade, photonics research has explored accelerated tensor operations, foundational to artificial intelligence (AI) and deep learning1,2,3,4, as a path towards enhanced energy efficiency and performance5,6,7,8,9,10,11,12,13,14. The field is centrally motivated by finding alternative technologies to extend computational progress in a post-Moore’s law and Dennard scaling era15,16,17,18,19. Despite these advances, no photonic chip has achieved the precision necessary for practical AI applications, and demonstrations have been limited to simplified benchmark tasks. Here we introduce a photonic AI processor that executes advanced AI models, including ResNet3 and BERT20,21, along with the Atari deep reinforcement learning algorithm originally demonstrated by DeepMind22. This processor achieves near-electronic precision for many workloads, marking a notable entry for photonic computing into competition with established electronic AI accelerators23 and an essential step towards developing post-transistor computing technologies.

This is a preview of subscription content, access via your institution

Access options

Access Nature and 54 other Nature Portfolio journals

Get Nature+, our best-value online-access subscription

$32.99 / 30 days

cancel any time

Subscribe to this journal

Receive 51 print issues and online access

$199.00 per year

only $3.90 per issue

Buy this article

USD 39.95

Prices may be subject to local taxes which are calculated during checkout

Additional access options:

Similar content being viewed by others

Data availability

The datasets presented in this study and analysis programs are available at https://github.com/lightmatter-ai/upaia-paper-2025.

References

  1. LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015).
    Article ADS CAS PubMed Google Scholar
  2. Krizhevsky, A., Sutskever, I. & Hinton, G. E. ImageNet classification with deep convolutional neural networks. Commun. ACM 60, 84–90 (2017).
    Article Google Scholar
  3. He, K., Zhang, X., Ren, S. & Sun, J. in Proc. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 770–778 (IEEE, 2016).
  4. Vinyals, O. et al. Grandmaster level in StarCraft II using multi-agent reinforcement learning. Nature 575, 350–354 (2019).
    Article ADS CAS PubMed Google Scholar
  5. Shen, Y. et al. Deep learning with coherent nanophotonic circuits. Nat. Photonics 11, 441–446 (2017).
    Article ADS CAS Google Scholar
  6. Lin, X. et al. All-optical machine learning using diffractive deep neural networks. Science 361, 1004–1008 (2018).
    Article ADS MathSciNet CAS PubMed Google Scholar
  7. Hamerly, R., Bernstei, L., Sludds, A., Soljačić, M. & Englund, D. Large-scale optical neural networks based on photoelectric multiplication. Phys. Rev. X 9, 021032 (2019).
    CAS Google Scholar
  8. Feldmann, J. et al. Parallel convolutional processing using an integrated photonic tensor core. Nature 589, 52–58 (2021).
    Article ADS CAS PubMed Google Scholar
  9. Dong, B. et al. Partial coherence enhances parallelized photonic computing. Nature 632, 55–62 (2024).
    Article CAS PubMed PubMed Central Google Scholar
  10. Dong, B. et al. Higher-dimensional processing using a photonic tensor core with continuous-time data. Nat. Photonics 17, 1080–1088 (2023).
    Article ADS CAS Google Scholar
  11. Becker, S., Englund, D. & Stiller, B. An optoacoustic field-programmable perceptron for recurrent neural networks. Nat. Commun. 15, 3020 (2024).
    Article ADS CAS PubMed PubMed Central Google Scholar
  12. Chen, Z. et al. Deep learning with coherent VCSEL neural networks. Nat. Photonics 17, 723–730 (2023).
    Article ADS CAS Google Scholar
  13. Wang, T. et al. An optical neural network using less than 1 photon per multiplication. Nat. Commun. 13, 123 (2022).
    Article ADS CAS PubMed PubMed Central Google Scholar
  14. Sludds, A. et al. Delocalized photonic deep learning on the internet’s edge. Science 378, 270–276 (2022).
    Article ADS CAS PubMed Google Scholar
  15. Shalf, J. The future of computing beyond Moore’s Law. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 378, 20190061 (2020).
    Article ADS MathSciNet Google Scholar
  16. Schwierz, F. & Liou, J. J. in Proc. 2020 IEEE Latin America Electron Devices Conference (LAEDC) 1–4 (IEEE, 2020).
  17. Leiserson, C. E. et al. There’s plenty of room at the Top: what will drive computer performance after Moore’s law? Science 368, eaam9744 (2020).
    Article CAS PubMed Google Scholar
  18. Moore, G. E. Cramming more components onto integrated circuits. Proc. IEEE 86, 82–85 (1998).
    Article Google Scholar
  19. Waldrop, M. M. The chips are down for Moore’s law. Nature 530, 144–147 (2016).
    Article ADS CAS PubMed Google Scholar
  20. Vaswani, A. et al. in Proc. Advances in Neural Information Processing Systems 30 (eds Guyon, I. et al.) 5998–6008 (Curran Associates, 2017).
  21. Devlin, J., Chang, M.-W., Lee, K. & Toutanova, K. in Proc. 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) (eds Burstein, J., Doran, C. & Solorio, T.) 4171–4186 (Association for Computational Linguistics, 2019).
  22. Mnih, V. et al. Human-level control through deep reinforcement learning. Nature 518, 529–533 (2015).
    Article ADS CAS PubMed Google Scholar
  23. Jouppi, N. P. et al. in Proc. 44th Annual International Symposium on Computer Architecture (ISCA ’17) 1–12 (ACM, 2017).
  24. Peng, B., Hua, S., Su, Z., Xu, Y. & Shen, Y. in Proc. 2022 IEEE Photonics Conference (IPC) (IEEE, 2022).
  25. Youngblood, N. Coherent photonic crossbar arrays for large-scale matrix-matrix multiplication. IEEE J. Sel. Top. Quantum Electron. 29, 1–11 (2023).
    Article Google Scholar
  26. Zhang, H. et al. An optical neural chip for implementing complex-valued neural network. Nat. Commun. 12, 457 (2021).
    Article ADS CAS PubMed PubMed Central Google Scholar
  27. Wetzstein, G. et al. Inference in artificial intelligence with deep optics and photonics. Nature 588, 39–47 (2020).
    Article ADS CAS PubMed Google Scholar
  28. Demirkiran, C. et al. An electro-photonic system for accelerating deep neural networks. ACM J. Emerg. Technol. Comput. Syst. 19, 1–31 (2023).
    Article Google Scholar
  29. Pintus, P. et al. Integrated non-reciprocal magneto-optics with ultra-high endurance for photonic in-memory computing. Nat. Photonics 19, 54–62 (2025).
    Article CAS Google Scholar
  30. Shastri, B. J. et al. Photonics for artificial intelligence and neuromorphic computing. Nat. Photonics 15, 102–114 (2021).
    Article ADS CAS Google Scholar
  31. Xu, X. et al. 11 TOPS photonic convolutional accelerator for optical neural networks. Nature 589, 44–51 (2021).
    Article ADS CAS PubMed Google Scholar
  32. Jacob, B. et al. in Proc. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition 2704–2713 (IEEE, 2018).
  33. Courbariaux, M., Bengio, Y. & David, J.-P. Training deep neural networks with low precision multiplications. Preprint at https://arxiv.org/abs/1412.7024 (2015).
  34. Kirtas, M. et al. Mixed-precision quantization-aware training for photonic neural networks. Neural Comput. Appl. 35, 21361–21379 (2023).
    Article Google Scholar
  35. Basumallik, A. et al. Adaptive block floating-point for analog deep learning hardware. Preprint at https://arxiv.org/abs/2205.06287 (2022).
  36. Giewont, K. et al. 300-mm monolithic silicon photonics foundry technology. IEEE J. Sel. Top. Quantum Electron. 25, 1–11 (2019).
    Article Google Scholar
  37. Ghafarian, H. et al. A 9-bit, 45 mW, 0.05 mm2 source-series-terminated DAC driver with echo canceller in 22-nm CMOS for in-vehicle communication. IEEE Solid-State Circuits Lett. 4, 10–13 (2021).
    Article Google Scholar
  38. Yu, K. et al. in Proc. 2015 IEEE International Solid-State Circuits Conference - (ISSCC) Digest of Technical Papers https://doi.org/10.1109/isscc.2015.7063098 (IEEE, 2015).
  39. McCreary, J. L. & Gray, P. R. All-MOS charge redistribution analog-to-digital conversion techniques. I. IEEE J. Solid-State Circuits 10, 371–379 (1975).
    Article ADS Google Scholar
  40. Jang, M. et al. Design techniques for energy-efficient analog-to-digital converters. IEEE Open J. Solid-State Circuits Soc. 3, 145–161 (2023).
    Article Google Scholar
  41. Ramkaj, A. T. et al. A 5-GS/s 158.6-mW 9.4-ENOB passive-sampling time-interleaved three-stage pipelined-SAR ADC with analog-digital corrections in 28-nm CMOS. IEEE J. Solid-State Circuits 55, 1553–1564 (2020).
    Google Scholar
  42. Lagos, J. et al. A 10.1-ENOB, 6.2-fJ/conv.-step, 500-MS/s, ringamp-based pipelined-SAR ADC with background calibration and dynamic reference regulation in 16-nm CMOS. IEEE J. Solid-State Circuits 57, 1112–1124 (2022).
    Article ADS Google Scholar
  43. de Lima, T. F. et al. Noise analysis of photonic modulator neurons. IEEE J. Sel. Top. Quantum Electron. 26, 1–9 (2020).
    Article Google Scholar
  44. Karpathy, A. nanoGPT. GitHub https://github.com/karpathy/nanoGPT (2023).
  45. Wang, C. et al. Integrated lithium niobate electro-optic modulators operating at CMOS-compatible voltages. Nature 562, 101–104 (2018).
    Article ADS CAS PubMed Google Scholar
  46. Abel, S. et al. Large Pockels effect in micro- and nanostructured barium titanate integrated on silicon. Nat. Mater. 18, 42–47 (2019).
    Article ADS CAS PubMed Google Scholar
  47. Youngblood, N., Chen, C., Koester, S. J. & Li, M. Waveguide-integrated black phosphorus photodetector with high responsivity and low dark current. Nat. Photonics 9, 247–252 (2015).
    Article ADS CAS Google Scholar
  48. Parkhi, O. M., Vedaldi, A., Zisserman, A. & Jawahar, C. V. in Proc. 2012 IEEE Conference on Computer Vision and Pattern Recognition 3498–3505 (IEEE, 2012).

Download references

Acknowledgements

We would like to thank K. C. Buckenmaier, M. Gould, C. Ramey, B. Dobbie, S. McKenzie, O. Yildirim, J. Talmage and M. Todd for their early contributions to the development of the photonic processor. We would also like to thank C. McCarter, N. Dronen, M. Forsythe, T. Lazovich, L. Levkova, D. Walter and D. Widemann for the development and implementation of the ABFP format. Also, we thank C. Chan, P. Clark, S. Cyphers, L. Huang, E. Hein, A. Hussein, S. Iyer, T. Kenney, S. Lines, A. Romano, T. Sarvey and Y. Sanders for their early contributions to the development of the software framework.

Author information

Authors and Affiliations

  1. Lightmatter, Mountain View, CA, USA
    Sufi R. Ahmed, Reza Baghdadi, Mikhail Bernadskiy, Nate Bowman, Ryan Braid, Jim Carr, Chen Chen, Pietro Ciccarella, Matthew Cole, John Cooke, Kishor Desai, Carlos Dorta, Jonathan Elmhurst, Bryce Gardiner, Elliot Greenwald, Shashank Gupta, Parry Husbands, Brian Jones, Anthony Kopa, Ho John Lee, Arulselvan Madhavan, Adam Mendrela, Nicholas Moore, Lakshmi Nair, Aditya Om, Subie Patel, Rutayan Patro, Rob Pellowski, Esha Radhakrishnani, Sandeep Sane, Nicholas Sarkis, Joe Stadolnik, Mykhailo Tymchenko, Gongyu Wang, Kurt Winikka, Alexandra Wleklinski, Josh Zelman, Ritesh Jain, Ayon Basumallik, Darius Bunandar & Nicholas C. Harris
  2. OpenAI, San Francisco, CA, USA
    Richard Ho

Authors

  1. Sufi R. Ahmed
  2. Reza Baghdadi
  3. Mikhail Bernadskiy
  4. Nate Bowman
  5. Ryan Braid
  6. Jim Carr
  7. Chen Chen
  8. Pietro Ciccarella
  9. Matthew Cole
  10. John Cooke
  11. Kishor Desai
  12. Carlos Dorta
  13. Jonathan Elmhurst
  14. Bryce Gardiner
  15. Elliot Greenwald
  16. Shashank Gupta
  17. Parry Husbands
  18. Brian Jones
  19. Anthony Kopa
  20. Ho John Lee
  21. Arulselvan Madhavan
  22. Adam Mendrela
  23. Nicholas Moore
  24. Lakshmi Nair
  25. Aditya Om
  26. Subie Patel
  27. Rutayan Patro
  28. Rob Pellowski
  29. Esha Radhakrishnani
  30. Sandeep Sane
  31. Nicholas Sarkis
  32. Joe Stadolnik
  33. Mykhailo Tymchenko
  34. Gongyu Wang
  35. Kurt Winikka
  36. Alexandra Wleklinski
  37. Josh Zelman
  38. Richard Ho
  39. Ritesh Jain
  40. Ayon Basumallik
  41. Darius Bunandar
  42. Nicholas C. Harris

Contributions

S.R.A., R.Ba., N.B., R.Br., J.Co., C.C., P.C., J.Ca., K.D., C.D., J.E., B.G., E.G., S.G., R.H., R.J., B.J., A.K., A.Me., E.R., S.S., N.S., J.S., M.T., A.W., J.Z., D.B. and N.C.H. contributed to the design and development of the photonic processor hardware. M.B., A.B., A.O., M.C., P.H., A.Ma., N.M., L.N., S.P., R.Pa., R.Pe., K.W., G.W. and H.J.L. contributed to the design and development of the software stack for the photonic processor. All authors contributed to the manuscript.

Corresponding authors

Correspondence toAyon Basumallik, Darius Bunandar or Nicholas C. Harris.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature thanks Anthony Rizzo and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Cite this article

Ahmed, S.R., Baghdadi, R., Bernadskiy, M. et al. Universal photonic artificial intelligence acceleration.Nature 640, 368–374 (2025). https://doi.org/10.1038/s41586-025-08854-x

Download citation

This article is cited by

Associated content