Slides
Page 1
Photonics-AI Powering the shift into photonics connectivity in AI computing Vickram Vathulya President & CEO vickram.vathulya@sivers-semiconductors.com October 2024
Page 2
Photonics-AI Photonics-AI 4 16 Safe Harbor Disclosure This presentation has been prepared by Photonics-AI Semiconductors AB (“Photonics-AI," the "Company," "we," "us," "our" or similar terms) for informational purposes only and not for any other purpose. This presentation is not a prospectus and is not an offer to sell, nor a solicitation of an offer to buy, securities. This presentation and the accompanying oral presentation are made pursuant to Section 5(d) of the Securities Act of 1933, as amended (the "Securities Act"), and are intended solely for investors that are either “qualified institutional buyers” (as defined in Rule 144A under the Securities Act) or institutions that are institutional “accredited investors” (as defined in Rule 501 of Regulation D under the Securities Act), solely for the purposes of familiarizing such investors with the Company and determining whether such investors might have an interest in a securities offering contemplated by the Company. We are not making any offer of securities at this time and cannot accept any orders for securities at this time. Any offer of securities will only be made pursuant to an available exemption from registration under federal and state securities laws). No such registration statement has been filed as of the date of this presentation. Forward-Looking Statements This presentation and the accompanying oral presentation contain “forward-looking” statements that are based on the Company’s current expectations or forecasts of future events, rather than past, events and outcomes, and are not guarantees of future performance. Forward-looking statements include all statements other than statements of historical fact contained in this presentation, including information or predictions concerning the Company’s future business, results of operations, financial performance, plans and objectives (including all aspects of a potential transaction with the SPAC and potential future acquisitions), competitive position, market trends, and potential growth and market opportunities. In some cases, you can identify forward-looking statements by words such as “intends,” “estimates,” “predicts,” “potential,” “continues,” “anticipates,” “plans,” “expects,” “believes,” “should,” “could,” “may,” “will”, "targets", "projects", “seeks” or the negative of these terms or other comparable terminology. Forward-looking statements involve risks and uncertainties, which may cause the Company’s actual results, performance or achievements to be materially different from those expressed or implied or otherwise contemplated by the forward-looking statements. Key factors that could cause actual results to be different than expected or anticipated include, but are not limited to: our goals and strategies; our future business development, financial condition and results of operations; our customer relationships and our ability to retain and expand our customer relationships; our ability to accurately predict our future revenues for the purpose of appropriately budgeting and adjusting our expenses; our ability to diversify our customer base and develop relationships in new markets; the level of demand in our customers’ end markets; our ability to attract, train and retain key qualified personnel; changes in trade policies, including the imposition of tariffs; results of operations and financial condition; regulatory developments in the United States and foreign countries; changes, unrest, or other disruptions in domestic and foreign business, ed within a registration statement the Companymarket, financial, political and legal conditions; our ability to protect our intellectual property rights; and other factors discussed in the “Risk Factors” section of the prospectus includ may file, as well as other documents the Company has filed, or may file, with the SEC. No such registration statement has been filed as of the date of this presentation. In light of these risks, uncertainties, and assumptions, the forward-looking events and outcomes discussed in this presentation may not occur and our actual results could differ materially and adversely from those expressed or implied in the forward-looking statements. You are cautioned not to place undue reliance on forward-looking statements, which are inherently unreliable and speak only as of the date of this presentation. The Company cannot guarantee, predict, or assure that the future results, performance, or events and circumstances described in the forward-looking statements will be achieved or occur. Many actual events and circumstances are beyond the control of the Company. The Company undertakes no obligation, and the Company expressly disclaims any obligation, to update or alter any forward-looking statements for any reason aft2er t7he date of this presentation, whether as a result of new information, future events or otherwise, except as required by law. In addition, statements that “we believe” and similar statements reflect the Company’s beliefs and opinions on the relevant subject solely as of the date of this presentation. These statements are based upon information available to the Company as of the date of this presentation whether or not identified in this presentation, and while the Company believes such information forms a reasonable basis for such statements, such information may be limited or incomplete, and such statements should not be read to indicate that the Company has conducted an exhaustive inquiry into, or review of, all potentially available relevant information. These statements are inherently uncertain, and you are cautioned not to unduly rely upon these statements. This presentation also contains estimates, projections and other statistical data made by independent parties and by the Company relating to, among others, market size and growth and other data about the Company’s industry and its business. These data involve a number of assumptions and limitations, are for illustrative purposes only, are inherently uncertain and are subject to a wide variety of significant business, economic, competitive and other risks and uncertainties, and you are cautioned not to give undue weight to such data. The Company has not independently verified the accuracy and completeness of the information obtained by third parties included in this presentation. Actual results may differ materially from the results expressed, implied, or otherwise contemplated by the estimates, projections, and other statistical data contained in this presentation, and the inclusion of such information in this presentation should not be regarded as a representation by any person that any such results will be achieved.
Page 5
Photonics-AI Interconnect is the Critical Bottleneck to Delivering AI at Scale 2 AI AI models continue to proliferate, grow and increase in technical complexity at an unpreceded pace (e.g., large language models) Compute Compute vendors continue to develop high performance processors to meet the demanding requirements of large- scale AI models Interconnect Despite rapid innovations in models and processors, the interconnect technology needed to scale AI infrastructure remains obsolete
Page 6
Photonics-AI Sivers Photonics-AI at a Glance We Design, Manufacture and Sell One of the World’s Most Advanced Custom Semiconductor Laser Arrays • Presently headquartered in Scotland and with offices in the United States. • Robust Organization ToSupport Continued Growth • Currently ~85 employees (12 PhDs) • Strong IP portfolio built over 25+ years • Proprietary Know-How and Trade Secrets • Fabrication Process Recipes • Chip Architectures, Epi wafer designs • Laser designs & Testing methodologies • 3 Patents granted (in last 3 years) • Further 16 Patents pending (in last 2 years) • Latest Patent Strategy focused on high-impact concepts around increasing laser array manufacturing yields 3
Page 7
Photonics-AI Photonics-AI is well-positioned to exploit major secular trends Engaged with leading suppliers in “momentum” markets 4 2028 SAM $1.1b1 Optical Sensing Consumer Continuous Biometrics Optical High-Speed Connectivity AI - High Performance Computing and Optical Datacom 1: The Optical Penetration and SAM are based solely on Company estimates as of August 2023 and are subject to risks and assumptions which could cause actual results to vary from such SAM estimates. The estimated SAM does not represent a prediction or guarantee of the actual SAM for the Company's products/business. Further information in slide 11. 2: InP 2022, Yole Intelligence, October 2022 2027 SAM $159m2 • Addressing Large SAMs in • AI Data Center $1,140m in 2028 • Non-invasive Biosensing $159m in 2027
Page 8
Photonics-AI Industry Leaders Agree that Optical Interconnect is Critical for the Success of AI 5 “Over the past decade, NVIDIA-accelerated computing has delivered a million-X speedup in AI. The next million-X will require new, advanced technologies like optical I/O to support the bandwidth, power and scale requirements of future AI and ML workloads and system architectures.” Rob Ober Chief Platform Architect for Data Center Products NVIDIA “Interconnect bandwidth during scale-out is critical to preventing the accelerators from stalling while waiting on network transfers for either data or gradients.” Christopher Berner Head of Compute OpenAI “…As an analogy, replacing electrical I/O with optical I/O in CPUs and GPUs to transfer data is like going from using horse-drawn carriages to distribute goods, limited in capacity and range, to using cars and trucks that can deliver much larger quantities of goods over much longer distances…” Press release June 2024; “Intel Demonstrates First Fully Integrated Optical I/O Chiplet” “The biggest companies in the world are hitting an energy power wall and experiencing massive challenges with AI scalability. Traditional chips push the boundaries of what’s possible to cool, and data centers produce increasingly large energy footprints. AI advances will slow significantly unless we deploy a new solution in data centers Nick Harris CEO and founder Lightmatter
Page 9
Photonics-AI6 Existing Copper Interconnect Technology Photonics-AI Creates Significant Power Efficiencies Using Light – deployed in Remote Laser Sources powering Co- Packaged Optics AI models are extremely energy intensive Datacenters are on track to be 20-50% of global electricity production by 2030 Optical I/O is the Solution for the Generative AI Era Photonics-AI is the solution Copper Photonics Comparison Power Consumption 50pJ/bit 5pJ/bit* 10X lower1 Latency 100ns 5ns Faster learning Cost $X $0.1X 90% reduction Network Size Scalability 10 meters 2,000 meters Larger clusters • 800G pluggable limited to 15pJ/bit 1. Source: https://ayarlabs.com//
Page 10
Photonics-AI Significant Momentum Backed by Marquee Customer Development Contracts 7 Ayar SuperNovaTM multi-port, multi-wavelength light source Photonics-AI laser AI Infrastructure Consumer F100’s Advanced Platform for Biometric Sensing Ayar Labs Investors ➢ ~$18 m USD continued multi year NRE ➢ More than 30,000 chips shipped Bi-directional data connectivity GPU TeraPHYTM CMOS Optical Chiplets
Page 11
Photonics-AI Photonics-AI’s Differentiation Moat 8 Photonics-AI has an established track- record of working closely with tech companies to deliver Custom Photonics Solutions (e.g. Ayar Labs and F100 customer) Advanced Performance Highly scalable integrated platform. A leading supplier offering Tunable Laser Arrays with Integrated Optical Power Monitors Powering advanced co-packaged optics interconnect technologies delivering 10x higher bandwidth at 90% lower power Demonstrated Skillset Trade secrets developed over 25 years and custom in-house design software deliver unmatched customizability for customers Strong design IP and patent portfolio to deliver industry leading test yields and strong gross margins Customer Benefit: ✓ Fastest time to market ✓ Higher capacity and lower cost Customer Benefit: ✓ Enabling differentiation for our customers ✓ Greater support, resources and expertise Customer Benefit: ✓ High channel density allowing more efficient use of available spectrum. ✓ Greater bandwidth scalability. Customer Intimacy
Page 12
Photonics-AI Photonics-AI Fab Photonics-AI’s Flexible Competitive and Asset-Light Manufacturing Strategy 9 Stage 1 Laser Design Semiconductor laser chip designs are created for either specific or general product usage. Stage 2 Wafer & Chip Manufacturing Semiconductor wafers are processed through a complex and extensive series of manufacturing steps including test and singulation. Stage 3 Assembly and Packaging Laser chips are assembled into packages to form the electronic components that can be mounted onto circuit boards. Customers Highly Scalable Infrastructure Limited investment required to add significant capacity Foundry for high volume • Unique integrated capability for laser design and fabrication. • $17M Capex investment expanding wafer fab capable of $150M/yr revenue1 • External foundry engagement underway for higher volumes Photonics-AI Photonics-AI 1. Based solely on Company estimate as of August 2024 and is subject to assumptions and uncertainties which could cause actual results to vary from the estimate and such estimates are not guarantees or predictions of actual performance.
Page 13
Photonics-AI 2022A 2023A 2024E 2025E 2026E 2027E 2028E Photonics-AI’s Serviceable Addressable Market Reaching $1.1B+ by 2028 11 Number of server GPU units shipped to grow from 4.7m in 2023 to 19.0m in 2028 (CAGR 32%) Support for large GPU clusters Size: 5-50k GPUs per cluster • 16Tb/s bi-directional connectivity per GPU- GPU link, • Ten laser arrays required per GPU Optical penetration: adoption rate based on assumptions driven by bottleneck factors in previous slide - increase in model size and pressure to reduce energy consumption 1. Source: IDC, Worldwide PC, Workstation, and Server Discrete Graphics Processing Unit Market Shares and Market Forecast, June 2024. 2. https://www.linkedin.com/posts/ayar-labs_aihwedgeaisummit-edgeai-aihardware-activity-7237978927896412161-PDLR?utm_source=share&utm_medium=member_desktop 3. The Optical Penetration and SAM are based solely on Company estimates as of August 2024 and are subject to risks and assumptions which could cause actual results to vary from such SAM estimates. The estimated SAM does not represent a prediction or guarantee of the actual SAM for the Company's products/business (SAM, $ Billions) Data Center GPU Units Sold (Millions)1 10.2 14.8 17.6 19.0 Laser Arrays per GPU2 10 10 10 10 Illustrative Optical Penetration 0.4% 2% 6% 12% Serviceable Addressable Market3 $31M $222M $528M $1,140 Serviceable Addressable Market for Photonics-AI Key Drivers Photonics-AI CW DFB Array 0.0 0.2 0.5 1.1 2025E 2026E 2027E 2028E
Page 14
AI Datacenter deployment: Socket Opportunity for Photonics-AI Assumptions: 1. Datacenter Power Utilization Efficiency (PUE) of 1.4 2. 60% of Datacenter IT power consumption by Server GPUs (others: storage, switches, networking etc.) 3. Assume 100% utilization of GPUs with Optical Interconnects and 2 suppliers for lasers at 50% share each 4. 10 Laser Arrays per GPU (actual multiplier will be based on customer configuration of photonics chiplets); 1000W GPU clusters500W GPU clusters $100M $50M 500 MW AI Datacenter 200 MW AI Datacenter $40M $20M
Page 15
Photonics-AI AI Eco-system from Chat GPT to light source (Photonics-AI) 12 Light sourcesOptical I/OGP/CPU vendorsServersCloud serviceAI Software tools Photonics-AI Illustrated is a sample of the AI Eco-system
Page 16
Photonics-AI 1. Estimated timeline only based on Company's expectations as of October 2024. Actual timeline is subject to change. 13 ✓ Technology Development ✓ Product Demonstration Product Qualification Manufacturing Readiness
Page 17
Photonics-AI Significant Funding and Valuations Throughout AI Photonics Ecosystem Key AI Photonics Fundraising Activity ▪ Developer of optical interconnects intended for data movement within AI systems ▪ Evolves alongside AI workloads and architectures, enabling customers to maximize the computing efficiency and performance of AI infrastructure while reducing costs, latency, and power consumption Notable Partnerships ▪ Data center and AI computing platform intended to serve deep learning and machine learning applications ▪ Combines the advantages of photonics, mixed-signalASICs, and packaging to offer a sustainable improvement in computing performance Notable Partnerships ▪ Integrated optical technology to create efficient processors and accelerates critical operations in neural networks using an array of programmable photonic elements fabricated alongside transistors in conventional CMOS processes Notable Partnerships Source: Pitchbook, Company websites. 16 36
Page 18
Thank you