![hataf_capital's tweet photo. Why Marvell $MRVL and Astera Labs $ALAB are the Stealth Architects of the AI Pivot
By Hataf Capital
The AI gold rush has a massive, unbilled expense that most investors are ignoring: the soaring cost of memory. While the spotlight remains fixed on GPU compute power, a quiet crisis is brewing in the supply chain. High Bandwidth Memory (HBM) and DRAM spot prices didn’t just tick upward last year they exploded four-fold between August and November. For the hyperscalers building the "brains" of the future, this isn't just a rounding error; it’s a direct assault on Cost of Goods Sold (COGS) and gross margins.
But in the semiconductor world, one man’s bottleneck is another’s windfall. As memory prices skyrocket, the industry is pivoting toward architectural efficiency. This shift is catapulting Marvell Technology $MRVL and Astera Labs $ALAB from peripheral players to essential architects of the AI data center.
Making "Old" Memory New Again
Hyperscalers like Microsoft $MSFT, Google $GOOG, and $META are facing a grim reality: they can no longer simply "brute force" their way through memory bottlenecks by throwing more expensive HBM at the problem. The math doesn't work when DRAM prices are up 400%.
Enter the Memory Controller.
Think of a memory controller as a high-speed traffic warden. Using the Compute Express Link (CXL) protocol, these chips allow data centers to pool and share memory across different CPUs and GPUs. More importantly, they give "second life" to older, cheaper iterations like DDR4. Instead of being forced into the latest, most expensive memory cycles, memory controllers allow systems to optimize existing resources without sacrificing the bandwidth required for heavy AI inference and recommendation models.
Marvell: The $2 Billion Stealth Play
Marvell recently posted Q3 results that finally showed the "inflection point" the market had been waiting for. While most of the noise was around their XPU and networking growth, the real story is their Structura CXL-based memory controller line.
Marvell’s management recently dropped a bombshell that the market seems to have missed: they expect their memory controller and NIC business to generate roughly $2 billion over the next three years. On an annualized basis, that’s ~$700M by 2028.
The analytical disconnect here is glaring. Current consensus estimates for Marvell’s 2026 revenue sit around $10B. My observation is that these estimates are largely "backward-looking." They account for the traditional networking business but fail to price in the margin-saving necessity of Structura as hyperscalers scramble to decouple their growth from volatile memory spot prices. If Marvell successfully scales this interoperability allowing $AMD EPYC and Intel $INTC Xeon chips to share memory seamlessly they aren't just selling a chip; they’re selling a solution to a multi-billion dollar procurement headache.
Astera Labs: The Pure-Play Winner with a Microsoft Edge
If Marvell is the diversified giant, Astera Labs is the agile specialist. For much of 2025, the bears focused on Astera's Aries retimer business, worried about creeping competition. They’re looking at the wrong rearview mirror.
Astera has been vocal about a $12B Total Addressable Market (TAM), where memory controllers represent a massive 33% slice. The "proof of concept" arrived two months ago when Astera locked in Microsoft Azure as a flagship customer for their Leo CXL memory controllers.
When a hyperscaler of Azure’s scale integrates your silicon to solve memory bottlenecks, you aren't "sampling" anymore you're foundational. By 2026, I expect switches and memory controllers to overtake retimers as Astera’s primary revenue drivers. Unlike the market, which is obsessing over 2025's competition, the real alpha lies in Astera’s ability to capture three-quarters of that $12B TAM through this technological pivot.
The Valuation Gap: Why the "Premium" is a Bargain
Data doesn't lie, but it does require context. Astera Labs trades at a forward revenue multiple of ~24x. To a value investor, that looks expensive. To an AI analyst, it looks like a steal when you realize revenues are projected to jump 109% to over $800M in 2026. You rarely get the chance to buy the leader of a niche, high-growth infrastructure segment at these levels before the full revenue ramp hits the balance sheet.
Marvell, meanwhile, trades at a more modest ~9.5x forward revenue. With a projected 22% revenue jump this year and a massive, unpriced tailwind from their memory controller guidance, the risk/reward profile is heavily skewed to the upside.
Final Verdict
The market is treating memory controllers like a "nice-to-have" accessory. They are wrong. As long as DRAM prices remain elevated, these components are a mission-critical survival tool for GPU makers and cloud providers alike.
Marvell $MRVL and Astera Labs $ALAB are not just "benefiting" from AI; they are providing the architectural escape hatch for companies drowning in memory costs. I am reiterating a Strong Buy on Astera Labs for its pure-play explosive growth and a Buy on Marvell for its massive, undervalued infrastructure footprint.
[Guys! Please Make Sure To Like And Repost If You Like Our Content]](https://pbs.twimg.com/media/G9_9KJ6bcAUZNOO.jpg)
Olivia
Online
ShahofDeccan
@ShahDean12
Joined February 2024
131 Following
37 Followers
68 Posts
@BalaamAndDonkey Balaam is options trading permissible according to Ismaili school?
@amaradunqas Was Hulegu Khan Shia? I think Oljaitu was Shia if im not wrong.
Even Hindus remember Imam Hussain but Nasibi sunnis don't. Never seen a sunni countries leader honor or remember the grandson of the Prophet.
The sacrifice of Hazrat Imam Hussain (AS) continues to inspire many people to remain steadfast in the pursuit of truth and justice. It is also a reminder of the enduring power of courage and conviction.
@Jalal8qh Jews atleast bought land in Palestine and gave equal citizenship to Palestinians, sunnis did none of that in Bahrain. Straight up demographic displacement and suppression.
Why Marvell $MRVL and Astera Labs $ALAB are the Stealth Architects of the AI Pivot
By Hataf Capital
The AI gold rush has a massive, unbilled expense that most investors are ignoring: the soaring cost of memory. While the spotlight remains fixed on GPU compute power, a quiet crisis is brewing in the supply chain. High Bandwidth Memory (HBM) and DRAM spot prices didn’t just tick upward last year they exploded four-fold between August and November. For the hyperscalers building the "brains" of the future, this isn't just a rounding error; it’s a direct assault on Cost of Goods Sold (COGS) and gross margins.
But in the semiconductor world, one man’s bottleneck is another’s windfall. As memory prices skyrocket, the industry is pivoting toward architectural efficiency. This shift is catapulting Marvell Technology $MRVL and Astera Labs $ALAB from peripheral players to essential architects of the AI data center.
Making "Old" Memory New Again
Hyperscalers like Microsoft $MSFT, Google $GOOG, and $META are facing a grim reality: they can no longer simply "brute force" their way through memory bottlenecks by throwing more expensive HBM at the problem. The math doesn't work when DRAM prices are up 400%.
Enter the Memory Controller.
Think of a memory controller as a high-speed traffic warden. Using the Compute Express Link (CXL) protocol, these chips allow data centers to pool and share memory across different CPUs and GPUs. More importantly, they give "second life" to older, cheaper iterations like DDR4. Instead of being forced into the latest, most expensive memory cycles, memory controllers allow systems to optimize existing resources without sacrificing the bandwidth required for heavy AI inference and recommendation models.
Marvell: The $2 Billion Stealth Play
Marvell recently posted Q3 results that finally showed the "inflection point" the market had been waiting for. While most of the noise was around their XPU and networking growth, the real story is their Structura CXL-based memory controller line.
Marvell’s management recently dropped a bombshell that the market seems to have missed: they expect their memory controller and NIC business to generate roughly $2 billion over the next three years. On an annualized basis, that’s ~$700M by 2028.
The analytical disconnect here is glaring. Current consensus estimates for Marvell’s 2026 revenue sit around $10B. My observation is that these estimates are largely "backward-looking." They account for the traditional networking business but fail to price in the margin-saving necessity of Structura as hyperscalers scramble to decouple their growth from volatile memory spot prices. If Marvell successfully scales this interoperability allowing $AMD EPYC and Intel $INTC Xeon chips to share memory seamlessly they aren't just selling a chip; they’re selling a solution to a multi-billion dollar procurement headache.
Astera Labs: The Pure-Play Winner with a Microsoft Edge
If Marvell is the diversified giant, Astera Labs is the agile specialist. For much of 2025, the bears focused on Astera's Aries retimer business, worried about creeping competition. They’re looking at the wrong rearview mirror.
Astera has been vocal about a $12B Total Addressable Market (TAM), where memory controllers represent a massive 33% slice. The "proof of concept" arrived two months ago when Astera locked in Microsoft Azure as a flagship customer for their Leo CXL memory controllers.
When a hyperscaler of Azure’s scale integrates your silicon to solve memory bottlenecks, you aren't "sampling" anymore you're foundational. By 2026, I expect switches and memory controllers to overtake retimers as Astera’s primary revenue drivers. Unlike the market, which is obsessing over 2025's competition, the real alpha lies in Astera’s ability to capture three-quarters of that $12B TAM through this technological pivot.
The Valuation Gap: Why the "Premium" is a Bargain
Data doesn't lie, but it does require context. Astera Labs trades at a forward revenue multiple of ~24x. To a value investor, that looks expensive. To an AI analyst, it looks like a steal when you realize revenues are projected to jump 109% to over $800M in 2026. You rarely get the chance to buy the leader of a niche, high-growth infrastructure segment at these levels before the full revenue ramp hits the balance sheet.
Marvell, meanwhile, trades at a more modest ~9.5x forward revenue. With a projected 22% revenue jump this year and a massive, unpriced tailwind from their memory controller guidance, the risk/reward profile is heavily skewed to the upside.
Final Verdict
The market is treating memory controllers like a "nice-to-have" accessory. They are wrong. As long as DRAM prices remain elevated, these components are a mission-critical survival tool for GPU makers and cloud providers alike.
Marvell $MRVL and Astera Labs $ALAB are not just "benefiting" from AI; they are providing the architectural escape hatch for companies drowning in memory costs. I am reiterating a Strong Buy on Astera Labs for its pure-play explosive growth and a Buy on Marvell for its massive, undervalued infrastructure footprint.
[Guys! Please Make Sure To Like And Repost If You Like Our Content]
![hataf_capital's tweet photo. Why Marvell $MRVL and Astera Labs $ALAB are the Stealth Architects of the AI Pivot
By Hataf Capital
The AI gold rush has a massive, unbilled expense that most investors are ignoring: the soaring cost of memory. While the spotlight remains fixed on GPU compute power, a quiet crisis is brewing in the supply chain. High Bandwidth Memory (HBM) and DRAM spot prices didn’t just tick upward last year they exploded four-fold between August and November. For the hyperscalers building the "brains" of the future, this isn't just a rounding error; it’s a direct assault on Cost of Goods Sold (COGS) and gross margins.
But in the semiconductor world, one man’s bottleneck is another’s windfall. As memory prices skyrocket, the industry is pivoting toward architectural efficiency. This shift is catapulting Marvell Technology $MRVL and Astera Labs $ALAB from peripheral players to essential architects of the AI data center.
Making "Old" Memory New Again
Hyperscalers like Microsoft $MSFT, Google $GOOG, and $META are facing a grim reality: they can no longer simply "brute force" their way through memory bottlenecks by throwing more expensive HBM at the problem. The math doesn't work when DRAM prices are up 400%.
Enter the Memory Controller.
Think of a memory controller as a high-speed traffic warden. Using the Compute Express Link (CXL) protocol, these chips allow data centers to pool and share memory across different CPUs and GPUs. More importantly, they give "second life" to older, cheaper iterations like DDR4. Instead of being forced into the latest, most expensive memory cycles, memory controllers allow systems to optimize existing resources without sacrificing the bandwidth required for heavy AI inference and recommendation models.
Marvell: The $2 Billion Stealth Play
Marvell recently posted Q3 results that finally showed the "inflection point" the market had been waiting for. While most of the noise was around their XPU and networking growth, the real story is their Structura CXL-based memory controller line.
Marvell’s management recently dropped a bombshell that the market seems to have missed: they expect their memory controller and NIC business to generate roughly $2 billion over the next three years. On an annualized basis, that’s ~$700M by 2028.
The analytical disconnect here is glaring. Current consensus estimates for Marvell’s 2026 revenue sit around $10B. My observation is that these estimates are largely "backward-looking." They account for the traditional networking business but fail to price in the margin-saving necessity of Structura as hyperscalers scramble to decouple their growth from volatile memory spot prices. If Marvell successfully scales this interoperability allowing $AMD EPYC and Intel $INTC Xeon chips to share memory seamlessly they aren't just selling a chip; they’re selling a solution to a multi-billion dollar procurement headache.
Astera Labs: The Pure-Play Winner with a Microsoft Edge
If Marvell is the diversified giant, Astera Labs is the agile specialist. For much of 2025, the bears focused on Astera's Aries retimer business, worried about creeping competition. They’re looking at the wrong rearview mirror.
Astera has been vocal about a $12B Total Addressable Market (TAM), where memory controllers represent a massive 33% slice. The "proof of concept" arrived two months ago when Astera locked in Microsoft Azure as a flagship customer for their Leo CXL memory controllers.
When a hyperscaler of Azure’s scale integrates your silicon to solve memory bottlenecks, you aren't "sampling" anymore you're foundational. By 2026, I expect switches and memory controllers to overtake retimers as Astera’s primary revenue drivers. Unlike the market, which is obsessing over 2025's competition, the real alpha lies in Astera’s ability to capture three-quarters of that $12B TAM through this technological pivot.
The Valuation Gap: Why the "Premium" is a Bargain
Data doesn't lie, but it does require context. Astera Labs trades at a forward revenue multiple of ~24x. To a value investor, that looks expensive. To an AI analyst, it looks like a steal when you realize revenues are projected to jump 109% to over $800M in 2026. You rarely get the chance to buy the leader of a niche, high-growth infrastructure segment at these levels before the full revenue ramp hits the balance sheet.
Marvell, meanwhile, trades at a more modest ~9.5x forward revenue. With a projected 22% revenue jump this year and a massive, unpriced tailwind from their memory controller guidance, the risk/reward profile is heavily skewed to the upside.
Final Verdict
The market is treating memory controllers like a "nice-to-have" accessory. They are wrong. As long as DRAM prices remain elevated, these components are a mission-critical survival tool for GPU makers and cloud providers alike.
Marvell $MRVL and Astera Labs $ALAB are not just "benefiting" from AI; they are providing the architectural escape hatch for companies drowning in memory costs. I am reiterating a Strong Buy on Astera Labs for its pure-play explosive growth and a Buy on Marvell for its massive, undervalued infrastructure footprint.
[Guys! Please Make Sure To Like And Repost If You Like Our Content]](https://pbs.twimg.com/media/G9_9VWEa8AAuSB4.jpg)
@OyeNawazShareef @MrAdnanRashid Jinnah was a parsing? Lmao 🤣 🤣 the cope! Jinnah was an ismaili shia who took mercy on you people and created a state which he shouldn't have done. You hateful rascals deserve to live under himdu boots and to get lynched for eating beef because of your hatred for others
@OyeNawazShareef @MrAdnanRashid Your country was founded by an ismaili shia you hateful retard. If you hate shias so much go to india
@DanzTrader That makes sense. Thanks!
@DGretta_Author What are the sources you use for data?
@Art_If_Ficial @tszzl Synthetic meat. Or or..
A whole planet just for cows lol. Fully automated and run by AI
@StratsLabs Thanks! I still remember your Cummins $cmi analysis. You were the only one who posted beforehand. Following your analysis with great interest!
@mukund @Rainwontmiss Any recommendations?
@visionstonks I saw your posts. If I join your discord, is it for lifetime? Also do you post entries and exits both or only entries?
Last Seen Users on Sotwe
jack
Seen from Germany
X
Seen from Denmark
Gizli Sokak Cekimi
Seen from Turkey
🧸
Seen from Saudi Arabia
يـــوســــف النــــيــاك 🍑🇧🇷
Seen from Switzerland
Moriyama Teshima Architects
Seen from United States
Majorboss 
Seen from United States
Mafouyz casa
Rose İfşa
Seen from Turkey
فضفضة واعترافات
Seen from Netherlands
Trends for you
Most Popular Users
Elon Musk
@elonmusk
241.5M followers
Barack Obama
@barackobama
119.1M followers
Cristiano Ronaldo
@cristiano
113.7M followers
Donald J. Trump
@realdonaldtrump
111.8M followers
Narendra Modi
@narendramodi
107.2M followers
Rihanna
@rihanna
98.5M followers
NASA
@nasa
92.4M followers
Justin Bieber
@justinbieber
91.7M followers
KATY PERRY
@katyperry
89.6M followers
Taylor Swift
@taylorswift13
83.5M followers
Lady Gaga
@ladygaga
75M followers
Virat Kohli
@imvkohli
72.7M followers
Kim Kardashian
@kimkardashian
70.7M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
65.6M followers
Bill Gates
@billgates
64.9M followers
Selena Gomez
@selenagomez
62.6M followers
The Ellen Show
@theellenshow
62.3M followers
CNN
@cnn
61.8M followers
X
@x
60.8M followers