1、 Technology|Semiconductors and Semiconductor Equipment Please see the analyst certification and important disclosures on page 25 of this report.Evercore ISI and affiliates do and seek to do business with companies covered in its research reports.Investors should be aware that the firm may have a con

2、flict of interest that could affect the objectivity of this report.Investors should consider this report as only a single factor in making their investment decision.2025.Evercore Group L.L.C.All rights reserved.DeepSeek Implications for Semis DeepSeeks released its R1 LLM on Jan 20th,which appears t

3、o lower the cost of training and inferencing by 10 x-30 x.We have 4 takes from this development:1)Lower Cost Compute Cycle=Increased Demand For Those Cycles.Historical case studies like virtualization of servers 20 years ago and Tectonic Shifts in computing over 70 years tell us that computing cost

4、improvements do not disrupt demand but rather translate to increased demand for lower cost compute-cycles.We expect this to manifest in greater proliferation of smaller-parameter set LLMs and spur the development of multi-mode(image/video)models that require higher computational intensity to train a

5、nd inference(positive for AI plays)2)Expect More AI in Smartphones and IoT Devices.We expect smaller LLM parameter sets will lead to accelerated deployment at the edge in Smartphones and IoT devices that will translate to applications that take advantage of low-latency,off-network computations(posit

6、ive for ARM,QCOM)3)More ASIC vs Merchant Debate.We expect the development to fuel increased debate around merchant vs ASIC solutions.History tells us that standard,general-purpose/flexible and programmable computing historically dominated previous computing eras,but this“parallel processing”era was