Technology / Semiconductor / Internet Research

Nicolas Baratte

Equity Bottom-Up

Micron Technology, Inc., through its subsidiaries, manufactures and markets dynamic random access memory chips (DRAMs), very fast static random access memory chips (SRAMs), Flash Memory, other semiconductor components, and memory modules.

<ul>
<li>Huge beat in Nov-25 and Feb-26. The impact of price hikes (Commodity Memory), mix change (more HBM and server) driven by AI growth and low capacity increase. </li>
<li>Management is cautious on bit growth but very positive on 2026-28 outlook. HBM sold out for ‘26. Shortages continue in ’26-27. HBM market to increase ~3x by ‘28.</li>
<li>Consensus should revise up FY26 by ~60%, FY27 by ~80%. This implies that the stock trades at 7.5x FY26 EPS and 5.4x FY27. </li>
</ul>


Huge beat in Nov-25 and Feb-26. The impact of price hikes (Commodity Memory), mix change (more HBM and server) driven by AI growth and low capacity increase. 
Management is cautious on bit growth but very positive on 2026-28 outlook. HBM sold out for ‘26. Shortages continue in ’26-27. HBM market to increase ~3x by ‘28.
Consensus should revise up FY26 by ~60%, FY27 by ~80%. This implies that the stock trades at 7.5x FY26 EPS and 5.4x FY27. 


Micron: Nov-25 Beat by 30%, Feb-26 Guidance 80% Above Consensus

NVIDIA Corporation designs, develops, and markets three dimensional (3D) graphics processors and related software.  The Company's products provide interactive 3D graphics to the mainstream personal computer market.

<ul>
<li>Who wins the AI race is like asking ”who won the Internet?” Amazon, Google, Meta, Microsoft? Wait, maybe it’s ASML and TSMC? Do you see how dumb this question is? </li>
<li>If you trade stocks, you enjoy the “Google is back” moment. Or maybe that was last week’s trade. The market wants profits and margins. Revenue growth is no longer enough.</li>
<li>So, who wins the AI race? I will eventually buy,  hopefully early 2026: AMD, Google, LAM, Meta, Microsoft, Micron, Nvidia, TSMC. These are the category winners IMO</li>
</ul>


Who wins the AI race is like asking ”who won the Internet?” Amazon, Google, Meta, Microsoft? Wait, maybe it’s ASML and TSMC? Do you see how dumb this question is? 
If you trade stocks, you enjoy the “Google is back” moment. Or maybe that was last week’s trade. The market wants profits and margins. Revenue growth is no longer enough.
So, who wins the AI race? I will eventually buy,  hopefully early 2026: AMD, Google, LAM, Meta, Microsoft, Micron, Nvidia, TSMC. These are the category winners IMO


Broadcom Inc. designs, develops, and supplies semiconductor and infrastructure software solutions. The Company offers storage adapters, controllers, networking processors, motion control encoders, and optical sensors, as well as infrastructure and security software to modernize, optimize, and secure the most complex hybrid environments. Broadcom serves customers worldwide.

<ul>
<li>Oracle: results were decent. Management released a huge Remaining Performance Obligation number aka future AI Infra revenue. That spooked the market. </li>
<li>Broadcom: a very good quarter, AI semi revenue guidance back on track. But Broadcom will reach its 2027 revenue target ($60-90bn) with 5 customers, not 3.</li>
<li>Results and outlook (for both) were ok to good; market reaction very negative.  1) the AI narrative is dead  2) the correction is especially steep for expensive stocks  </li>
</ul>


Oracle: results were decent. Management released a huge Remaining Performance Obligation number aka future AI Infra revenue. That spooked the market. 
Broadcom: a very good quarter, AI semi revenue guidance back on track. But Broadcom will reach its 2027 revenue target ($60-90bn) with 5 customers, not 3.
Results and outlook (for both) were ok to good; market reaction very negative.  1) the AI narrative is dead  2) the correction is especially steep for expensive stocks  


Oracle, Broadcom: Communication Errors – Bullish Tone when the Market Only Hear Concerns

<ul>
<li>Pres Trump authorized sales of H200 to China, with unknown conditions attached.  Not clear who the US will agree to sell to, who in China will be “allowed” to buy</li>
<li>More clear: H20 was not appealing – too under-powered. H200 is above the threshold at 2x the performance of Huawei Ascend 910C. So, Nvidia could find buyers in China.</li>
<li>H200 could represent 23% upside to Nvidia 2026 revenues (max). But the stock price moved up 1.7% on Monday, down -0.3% yesterday. The market is concerned by bigger issues.</li>
</ul>


Pres Trump authorized sales of H200 to China, with unknown conditions attached.  Not clear who the US will agree to sell to, who in China will be “allowed” to buy
More clear: H20 was not appealing – too under-powered. H200 is above the threshold at 2x the performance of Huawei Ascend 910C. So, Nvidia could find buyers in China.
H200 could represent 23% upside to Nvidia 2026 revenues (max). But the stock price moved up 1.7% on Monday, down -0.3% yesterday. The market is concerned by bigger issues.


Jensen Got What He Wanted And/Or Pres. Trump Needed a Sweetener for China

<ul>
<li>NVDA has $500bn revenue visibility for Blackwell and Rubin. This implies that Data Center Computing revenues will increase by ~70% in 2026.</li>
<li>Cloud capacity is fully utilized, 6-year old GPU are fully utilized. Don’t worry about depreciation theories. But the gap between capex and future revenues / profits will remain.</li>
<li>Consensus is ~20% too low for FY27 and possibly more for FY28, Stock trading below avg PEx.</li>
</ul>


NVDA has $500bn revenue visibility for Blackwell and Rubin. This implies that Data Center Computing revenues will increase by ~70% in 2026.
Cloud capacity is fully utilized, 6-year old GPU are fully utilized. Don’t worry about depreciation theories. But the gap between capex and future revenues / profits will remain.
Consensus is ~20% too low for FY27 and possibly more for FY28, Stock trading below avg PEx.


NVDA Strong Quarter, Strong Guidance, Consensus ~20% Too Low, Stock Is Not Expensive

Nicolas Baratte has been a financial analyst since 2000, covering the Semiconductor sector and Global Technology. Nicolas has worked on the sell-side for JPMorgan, CLSA. On the buy-side for a Tiger fund and Prudential. Nicolas has covered semiconductors, Asian hardware and components, and the Global Tech supply chain. Nicolas lives mostly in Taiwan, sometimes in France and is in New York regularly.

Nicolas Baratte
Technology / Semiconductor / Internet Research

Micron: Nov-25 Beat by 30%, Feb-26 Guidance 80% Above Consensus

strong performance in Nov-25 and Feb-26 due to price hikes and AI growth. Management optimistic for 2026-28 despite bit growth concerns. Stock...

Nicolas Baratte

Who’s Winning the AI Race?

The article questions the idea of "winning" the AI race and emphasizes the importance of profits and margins, suggesting potential category...

Nicolas Baratte

Oracle, Broadcom: Communication Errors – Bullish Tone when the Market Only Hear Concerns

Oracle and Broadcom had decent results, but market reacted negatively to Oracle's future AI revenue and Broadcom's reliance on 5 customers to...

Nicolas Baratte

Jensen Got What He Wanted And/Or Pres. Trump Needed a Sweetener for China

Trump authorized sales of H200 to China, potentially boosting Nvidia's revenue by 23%, but market remains concerned about larger issues.

Nicolas Baratte

NVDA Strong Quarter, Strong Guidance, Consensus ~20% Too Low, Stock Is Not Expensive

NVDA. Consensus estimates for FY27 are 20% too low, with potential for even more growth in FY28. Stock is trading below average PEx.

Nicolas Baratte

Micron: Nov-25 Beat by 30%, Feb-26 Guidance 80% Above Consensus

strong performance in Nov-25 and Feb-26 due to price hikes and AI growth. Management optimistic for 2026-28 despite bit growth concerns. Stock...

Nicolas Baratte

Who’s Winning the AI Race?

​The article questions the idea of "winning" the AI race and emphasizes the importance of profits and margins, suggesting potential category...

Nicolas Baratte

Oracle, Broadcom: Communication Errors – Bullish Tone when the Market Only Hear Concerns

​Oracle and Broadcom had decent results, but market reacted negatively to Oracle's future AI revenue and Broadcom's reliance on 5 customers to...

Nicolas Baratte

Jensen Got What He Wanted And/Or Pres. Trump Needed a Sweetener for China

​Trump authorized sales of H200 to China, potentially boosting Nvidia's revenue by 23%, but market remains concerned about larger issues.

Nicolas Baratte

NVDA Strong Quarter, Strong Guidance, Consensus ~20% Too Low, Stock Is Not Expensive

​NVDA. Consensus estimates for FY27 are 20% too low, with potential for even more growth in FY28. Stock is trading below average PEx.

Nicolas Baratte

The article questions the idea of "winning" the AI race and emphasizes the importance of profits and margins, suggesting potential category...

Oracle and Broadcom had decent results, but market reacted negatively to Oracle's future AI revenue and Broadcom's reliance on 5 customers to...

Trump authorized sales of H200 to China, potentially boosting Nvidia's revenue by 23%, but market remains concerned about larger issues.

NVDA. Consensus estimates for FY27 are 20% too low, with potential for even more growth in FY28. Stock is trading below average PEx.