AMD vs Intel Stock 2026: The Indian Investor's Complete Playbook (Foundry Thesis, LRS Math, Estate Tax)

It depends on whether you are buying execution or recovery. AMD trades at a higher forward price-to-earnings ratio of approximately 30 to 35 times because it is the perceived winner in AI accelerators outside Nvidia and in server CPU share gains against Intel. AMD has zero fabrication facilities and is asset-light, generating positive free cash flow of approximately 3 billion dollars on a trailing twelve month basis through mid 2025. Intel trades at approximately 25 to 30 times forward earnings but carries the weight of approximately 25 billion dollars in cumulative foundry losses since 2022 and an enormous capital expenditure program of 25 to 30 billion dollars annually. Intel free cash flow has been negative 9 to 12 billion dollars on a trailing twelve month basis. For an investor confident in Intel 18A node yields and the CHIPS Act funded foundry business, Intel is the higher-asymmetry recovery bet. For an investor seeking exposure to current AI compute and cloud server demand, AMD is the cleaner pure play.

May 2024 was the first time AMD market capitalization crossed Intel since the two companies have publicly traded. The drivers were three. First, AMD MI300 series AI accelerators ramped from near zero in 2023 to roughly 5 billion dollars annual revenue in 2024, taking the second-source position behind Nvidia. Second, AMD EPYC server CPUs took share from Intel Xeon, rising from under 5 percent of server CPU shipments in 2018 to approximately 30 percent by 2024 to 2025, with the AWS, Microsoft Azure and Google Cloud Tier 1 hyperscalers deploying EPYC at scale. Third, Intel foundry losses widened materially through 2022 to 2024 while Pat Gelsinger pushed the company through an aggressive capital expenditure cycle for the 18A node. Intel cut its dividend by 75 percent in August 2024 to 12.5 cents per quarter from 50 cents previously, signaling capital-allocation pivot. Pat Gelsinger departed as CEO in December 2024, replaced amid uncertainty about whether the foundry business would be spun off or retained.

Three routes. The Liberalised Remittance Scheme allows Indian residents to remit up to 250,000 dollars per financial year for investment purposes. Brokers like Vested, INDmoney, Groww US Stocks and IndianBoy offer fractional share purchase with all-in costs of approximately 1.2 to 1.8 percent including forex spread and brokerage. Interactive Brokers India offers cheaper execution at approximately 0.4 to 0.7 percent all-in but requires a higher minimum balance and full margin account documentation. The third route is GIFT City through India INX where US equity products are available but liquidity is thin and instrument selection limited. Tax Collected at Source of 20 percent applies on remittance above 10 lakh rupees per financial year, refundable as TDS credit at the time of filing your ITR. Schedule FA disclosure of all foreign assets is mandatory annually, with Black Money Act exposure for non-disclosure including 10 lakh rupee fines and possible imprisonment. The honest practical answer is Vested or INDmoney for under 5 lakh annual exposure, Interactive Brokers above that.

It is the largest single risk for Indian residents holding US-situs assets and almost never mentioned by Indian financial planners. US federal estate tax applies to non-resident decedents on US-situs assets above a 60,000 dollar threshold, at progressive rates from 18 percent to 40 percent. US-situs assets include directly-held US stocks like AMD and Intel listed on NASDAQ and NYSE. The estate tax is paid by the estate before assets are released to heirs, and the executor must file Form 706-NA. The India-US Double Taxation Avoidance Agreement does not contain an estate tax treaty, so no foreign tax credit relief is available. Practical mitigations include holding through Ireland-domiciled or Luxembourg-domiciled ETFs which are not US-situs assets, holding through a US-resident family trust structure which is complex and expensive, or simply staying below the 60,000 dollar US-situs exposure threshold. Most Indian investors with material US stock holdings have unmitigated estate tax exposure that surfaces only at the worst possible time.

Honest answer: own both if you must own either. Nvidia is the dominant AI chip supplier with over 80 percent market share in training compute, gross margins of approximately 75 percent in data center, and the only mature CUDA software ecosystem at scale. AMD MI300 series is the second-source AI accelerator with 5 billion dollars in 2024 revenue and 7 billion dollar plus 2025 guidance, but on a 90 billion dollar plus base for Nvidia data center revenue, AMD is roughly 8 to 10 percent the size in AI. For Indian investors with LRS access, a 60 percent Nvidia and 40 percent AMD split among AI compute names is more diversified than either alone. The thesis distinction is that Nvidia is priced for AI dominance to continue indefinitely while AMD is priced for share gains from a small base. Either thesis can be right. Combining both reduces single-name risk on what is currently the most concentrated trade in global equities.

It depends on Intel 18A node yields, which remain the single most important variable. Intel 18A is targeted to enter high-volume manufacturing in 2025 and is the technology bet that determines whether Intel becomes a credible foundry to TSMC or remains a captive design house. The bull case is that Intel 18A achieves yields above 70 percent within 12 to 18 months of ramp, attracts external foundry customers including potentially Apple and Qualcomm, and the foundry business becomes a separately-valued asset. The bear case is that 18A yields disappoint, customer adoption lags, foundry losses persist into 2027, and the foundry is eventually sold or spun off at a fraction of book value. The current Intel stock price implies a probability roughly midway between these scenarios. Position sizing should reflect this binary asymmetry, with Intel held at no more than 3 to 5 percent of portfolio for most retail investors.

Material and structural. AWS Graviton is Amazon's in-house ARM-based server CPU, designed by Annapurna Labs which AWS acquired in 2015. By 2025, Graviton instances account for approximately 50 percent of all new EC2 capacity additions at AWS. Google Cloud is rolling out Axion ARM CPUs in 2025. Microsoft Azure announced Cobalt 100 ARM CPUs in late 2023 deploying through 2025. Each ARM-based custom CPU instance is a transaction that AMD and Intel did not make. Cumulative impact through 2027 could be 8 to 12 percent of total cloud server CPU revenue diverted from x86 to ARM. The defensive argument for AMD is that AWS Graviton, Google Axion and Microsoft Cobalt are TSMC-fabricated, so AMD as a fabless TSMC customer is not structurally disadvantaged. The defensive argument for Intel is that 18A foundry could fabricate competitor designs and capture some of the value. Neither argument fully offsets the share loss. For Indian investors comparing AMD versus Intel, the custom-silicon threat affects both but Intel is more exposed because its fabs need to be filled.

Lisa Su became AMD CEO in October 2014 when the stock traded near 3 dollars and the company faced existential risk after years of share loss to Intel and inability to compete in server CPUs. Under her leadership, AMD launched the Zen microarchitecture in 2017 with Ryzen client CPUs and EPYC server CPUs, reversed market share losses to Intel, completed the 49 billion dollar Xilinx acquisition in 2022 expanding into FPGA and adaptive computing, launched MI300 AI accelerators competitive with Nvidia, and crossed Intel in market capitalization in May 2024. Stock returned approximately 100 times from her start. The case study is taught at business schools as an execution-driven turnaround that combined a focused technology bet on chiplet architecture, a strategic shift to TSMC for advanced node manufacturing, and disciplined capital allocation through acquisitions that strengthened rather than diluted the core thesis. Lisa Su remains CEO in 2026, which is itself a stability premium investors price into AMD versus Intel's CEO churn.

Both. The semiconductor industry has clear 3 to 5 year cycles driven by data center capex, PC refresh cycles, and inventory corrections. The SOX index has experienced 30 to 60 percent drawdowns in 2008, 2018, 2022, and any future downturn will likely catch AMD and Intel together. Within this cyclicality is a secular trend of rising silicon content per device, AI compute demand growth, and physical infrastructure modernization that grows total addressable market over decades. The correct framing is secular trend with cyclical entries. Historical data shows that buying AMD or Intel at peak forward earnings revisions and peak gross margin assumptions has produced negative real returns for 5 to 10 year holding periods. Buying at cycle bottoms when consensus pessimism aligns with inventory write-downs has produced 3 to 10 times returns in the subsequent cycle. AMD currently trades closer to mid-cycle valuations while Intel trades closer to a depressed multiple, reflecting their different cycle positions.

AMD has executed three transformative acquisitions in the last four years that fundamentally changed its addressable market and competitive position. Xilinx for 49 billion dollars closed in February 2022, adding FPGA and adaptive computing capability that expanded AMD into networking, automotive, aerospace and industrial markets. Pensando for 1.9 billion dollars in 2022 added Data Processing Unit capability for cloud and enterprise networking. Silo AI for 665 million dollars in 2024 added European AI software and inference expertise. ZT Systems for 4.9 billion dollars in 2024 added rack-scale AI server design and manufacturing capability, directly competing with Nvidia DGX and Supermicro. Cumulative spend exceeds 56 billion dollars in four years against AMD market cap that crossed 240 billion in 2024. The strategic thesis is that AMD is building a full-stack AI and computing solutions company rather than a chip-only design house. The execution risk is that integration of four acquisitions creates organizational complexity that has historically tripped up tech mergers. Stock performance through 2026 will reflect how well AMD digests these acquisitions.

Take a 1,000 dollar purchase of AMD as the worked example, equivalent to approximately 83,000 rupees at current INR exchange rate. Cost components. Forex spread of 0.5 to 1.0 percent on Vested or INDmoney, which is 415 to 830 rupees. Brokerage of 0 to 5 dollars per trade depending on platform, which is 0 to 415 rupees. TCS of 20 percent on the remittance amount above 10 lakh per financial year cumulative, refundable as TDS credit. For first-time small remitters below the 10 lakh threshold no TCS applies. Annual demat or platform fees of 0 to 2,000 rupees. For a small one-time purchase of 83,000 rupees below the TCS threshold, total effective drag is approximately 0.6 to 1.6 percent. For a larger 5 lakh rupee purchase that triggers TCS, the headline drag is 21 to 22 percent including TCS, but the TCS amount is fully refundable on ITR filing. Net of TCS refund the effective drag is approximately 0.6 to 1.6 percent. The hidden cost is opportunity cost of TCS sitting with the Income Tax department for 6 to 18 months before refund processes.

Neither should be overweight for most retail Indian investors. Both AMD and Intel are single-name positions in highly cyclical semiconductors, and even the bull thesis on either is sensitive to macro factors, AI capex trends, and execution surprises. Reasonable position sizing is no more than 3 to 5 percent of total portfolio in any single semiconductor stock including Nvidia. For diversified semiconductor exposure, the iShares Semiconductor ETF SOXX or VanEck Semiconductor ETF SMH bought through LRS captures AMD, Intel, Nvidia, TSMC ADR, Broadcom, Qualcomm and Marvell in one instrument at 0.35 percent expense ratio. The behavioral risk of overweighting AMD or Intel is recency bias from the 2023-2025 AI rally. Position size such that a 30 to 50 percent drawdown, which is consistent with historical sector corrections, would not force you to sell at the bottom. For broader US stock cost mechanics see the LRS true cost breakdown.