Jabil looks like a compelling long at $38B because markets have not priced in its 1.6T LRO pluggable transceiver business. With Sivers as the H1 2027 bottleneck, the question is how much it can make, not whether demand exists. It has massive supply chains, took over Intel's pluggable lines and looks more scalable than Applied Optoelectronics' ATM-funded laser fabs. Recognition likely comes in H1 2027; a ~40% rerating seems plausible, though at its size tripling is less likely. x.com  x.com 

Celestica's cheap TPU exposure offers upside, but its low margins and smaller moat than Vicor's make it a trade rather than a large core holding. It will probably reach $400-$500 over time; $400 next year looks likely. An attached analyst note expects scrutiny of a cut in 2026 free-cash-flow guidance from $600M to $460M due mainly to working capital; it warns that 2027 free cash flow could turn negative if working-capital investment stays high alongside $1.5B-$2B of capex. x.com  x.com  x.com 

Jabil's valuation already prices in most of its AI tailwind, so the current price is a chase rather than a buy. Intelligent Infrastructure revenue jumped 33.9% to $12.32B on AI server and liquid-cooling builds, and Jabil is partnering with Apptronik to manufacture the Apollo humanoid. At 17.1x forward earnings and 12.3x EV/EBITDA it looks reasonable, but dips below about $300 are the buy zone, not $338. x.com 

Celestica's Q3 beat and first 2027 guidance make its growth look underpriced. Revenue rose 83% and EPS 109%; management raised 2026 revenue guidance to $21B and put 2027 revenue at $40B-plus and EPS at $21.50, implying about 90% and 85% growth. At roughly 12x 2027 EPS after hours, shares are probably 50% undervalued; initial targets may have upside, with revenue growth potentially reaching 100%. At $34.50 of 2028 EPS and 24x earnings, shares would be worth $828. x.com  x.com  x.com  x.com  x.com 

Jabil's photonics business is underpriced, as the market values it as a contract manufacturer shipping 1.6T transceivers while its photonics revenue tells a different story. Per @aleabitoreddit, Jabil looks compelling at $38B, with massive supply chains in place, Intel's former pluggable lines and a validated hyperscaler supply chain, a more scalable model than building laser fabs. With Sivers as the bottleneck into H1 2027, the setup looks very interesting. x.com 

Celestica's valuation is too high if its expected growth over the next three years does not materialize. Contract electronics manufacturers usually trade at low-teens earnings multiples in normal times and single digits in bear markets, as Jabil shows. A reply argues it trades at 36x 2026 earnings on 50% growth and would be under 25x 2027 earnings, but that rests on the growth arriving. x.com 

Vicor's accelerating business and operating leverage make it look like an incredible winner. Revenue and operating margins jumped in the latest quarter; CEO Patrizio Vinciarelli said two additional sites would enable nearly 5x Fab-1's capacity across three fabs, citing demand and industry trends. But QQQ could fall to 660 before a meaningful bounce, so Vicor is not a buy in this tape. x.com  x.com 

Marvell builds the chips for the data-center pipes that AI keeps choking, and its numbers now show that thesis. Last quarter's revenue grew 36%. At 40 times forward earnings the stock is no longer cheap, so further gains must come from earnings growth. x.com 

CoreWeave's results confirm the AI supercycle, and the next wave of capital is moving beyond semiconductors after their big runs. CoreWeave reported $2.58 billion of revenue, up 122%, a $104 billion backlog and more than $25 billion of new customer commitments already in Q3. That capital is flowing into data centers (Nebius, IREN, Riot, Applied Digital), networking (Arista, Marvell), photonics (Lumentum, Coherent, Applied Optoelectronics) and power (Vertiv, Constellation, Vistra, Oklo). x.com  … Show more

Vertiv should beat and keep raising 2026 guidance because AI data center demand still exceeds supply. The proprietary pipeline added 50+ GW in 1Q26, 415 GW in the US since January 2023. Even flat orders near $18 billion would lift backlog from $15 billion to about $18 billion, implying about $25 billion of 2027 sales versus the Street's $16.9 billion. Base EPS of $6.75/$9.50 for 2026/27 tops the Street at a PEG under 1x. Margins are the watch item; the big run makes the print less attractive. … Show more