User's blog

Author: Antonis Kazoulis

  • Nvidia’s Kyber Rack Delayed to 2028 Following Manufacturing Challenges

    Nvidia’s Kyber Rack Delayed to 2028 Following Manufacturing Challenges

    YWO NEWS | EQUITIES | DEEP DIVE / ANALYSIS

    Nvidia’s next-generation Kyber NVL144 rack architecture has been pushed back to 2028 after manufacturing difficulties with a key circuit board proved insurmountable on the original timeline, research firm SemiAnalysis reported on Monday, CNBC reported. The system had been slated to debut alongside Vera Rubin Ultra in 2027, making this a slip of more than 12 months for what Nvidia had positioned as the centrepiece of its next rack-scale generation.

    NVDA opened Monday’s premarket session last down less than 0.1% at $194.79, per CNBC data. The stock showed little movement in premarket trading following the report. .

    The PCB Midplane Problem

    Kyber is not a chip — it is a server cabinet that packs 144 of Nvidia’s most powerful GPUs into a single unit, wiring them to function as one integrated compute block. The system mounts chips vertically rather than horizontally to increase density and cut latency. At its heart sits a specialised multi-layer printed circuit board — the PCB midplane — that connects the electronic modules within the cabinet.

    That midplane is where the programme has stalled.

    “Kyber NVL144 rack architecture has been delayed to 2028 as the PCB midplane remains challenging from a manufacturability standpoint,” SemiAnalysis said, as reported by CNBC.

    The NVL576 — a larger configuration linking eight Kyber racks through optical connections — is also either delayed or constrained to small volumes, SemiAnalysis added, per the same CNBC report.

    Nvidia did not respond to CNBC’s request for comment, CNBC noted.

    The Backup Plan Was Already Dead

    What makes the timeline slip more consequential is what happened to the contingency. Nvidia had explored bridging the gap by bolting two current-generation racks together to approximate Kyber’s compute footprint. Cloud service providers and hyperscalers rejected the workaround outright.

    “It has since been cancelled due to heavy pushback from CSPs and hyperscalers over its odd design and heavy operational burden,” SemiAnalysis said, per CNBC.

    According to SemiAnalysis, the alternative design was not pursued further. . SemiAnalysis concluded that Nvidia now has “no proven solution to expand the scale-up world size for Rubin Ultra,” as CNBC reported, citing the research firm directly. The double-rack design was reportedly presented at Nvidia’s GTC conference in San Jose on 16 March 2026, where Jensen Huang appeared onstage next to a Vera Rubin Ultra Kyber Compute Tray and a Vera Rubin Ultra Kyber NVLink MidPlane.

    Where AMD and Google May Find a Gap

    SemiAnalysis flagged that the void at the top of the rack-scale market could give Advanced Micro Devices and Google, whose in-house chips have already been winning business from top AI labs, a technical opening they have not had before, according to CNBC.

    The framing matters. Nvidia has operated on a roughly annual cadence at the chip level, and competitors have struggled to match the release rhythm. SemiAnalysis noted that any delay to Nvidia’s rack-scale roadmap could be relevant to competitors operating in the same market segment. 

    That said, the current-generation picture remains intact. Nvidia’s existing Rubin systems are in full production and begin shipping this autumn to eight cloud partners, including Amazon Web Services, Microsoft Azure, and Google Cloud, CNBC reported. And SemiAnalysis projected Nvidia’s data-centre compute revenue will run 20% above Wall Street consensus in the second half of fiscal 2027 .

    Cadence Under Strain

    The Kyber slip is part of a pattern SemiAnalysis characterised as Nvidia’s annual release cadence colliding with manufacturing limits, per CNBC. Designing ever-denser rack architectures is one problem. Getting Taiwan’s manufacturing base to produce the specialised multi-layer PCBs at volume and yield is another.

    Nvidia’s chip roadmap and its rack-scale architecture roadmap are no longer moving in lockstep. The chip — Rubin Ultra — is on schedule for 2027. The cabinet designed to house it at full scale is now a year behind. For hyperscalers who have already committed capex plans around the Kyber timeline, that divergence may affect product deployment timelines for some customers. 

    The commercial impact of the reported delay remains uncertain based on publicly available information. . What SemiAnalysis and Anniek Bao’s CNBC report do establish is that Nvidia has no announced bridge product for the highest-density configurations and no confirmed revised timeline beyond the 2028 guidance from SemiAnalysis.

    The 20% data-centre revenue beat projected for the second half of fiscal 2027 may help offset some near-term operational concerns. However, the longer-term competitive impact of the reported delay remains uncertain. .

    Risk Disclaimer: Trading CFDs involves substantial risk and may not be suitable for all investors. CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. You may lose some or all of your invested capital. You should consider whether you understand how CFDs work and whether you can afford to take the high risk of losing your money. Past performance is not a reliable indicator of future results. This article is provided for general informational and educational purposes only and does not constitute financial, investment, legal, tax, or trading advice, nor a recommendation, solicitation, or offer to buy or sell any financial instrument.

  • Goldman Sachs Lifts USD/JPY Forecasts to 165, Citing Structural Dollar Support

    Goldman Sachs Lifts USD/JPY Forecasts to 165, Citing Structural Dollar Support

    Wall Street’s bullish dollar thesis leaves yen without a near-term recovery catalyst.

    Goldman Sachs raised its USD/JPY forecasts on Sunday, projecting the pair at 162 in three months, 163 in six months, and 165 in 12 months, up from earlier targets of 160, 158, and 155 respectively, CNBC’s Lee Ying Shan reported. The revision came after the yen fell to its weakest level against the dollar in four decades last week — a move that has kept Japan’s Ministry of Finance focused on the currency market and investors watching for any official response.

    The bank’s case is structurally driven, not tactical. Goldman cited “higher-for-longer U.S. yields, low recession risk, lingering fiscal concerns, and only gradual BoJ hikes” as the forces it expects to keep depreciation pressure on the yen sustained, per the CNBC report. These factors may contribute to continued focus on the interest-rate differential between the U.S. and Japan. 

    Intervention Won’t Hold the Line, Goldman Says

    Japan’s Ministry of Finance has intervened in yen markets before. Goldman’s read on what happens next time is blunt:

    “We see no reason for the upward trend in USD/JPY to stop without an unexpected negative US growth shock or a BoJ pivot towards more aggressive policy tightening,” Goldman said, as cited by CNBC.

    The bank added that previous interventions had only temporarily interrupted the yen’s slide before USD/JPY resumed climbing, and it expects a similar pattern if authorities step in again. Goldman reportedly suggested that previous interventions had a temporary effect, although future outcomes may differ depending on market and policy conditions. .

    Japan’s own fiscal policy may compound the problem. Goldman flagged that Japan’s domestic stimulus plans could push up Japanese government bond term premiums relative to U.S. Treasurys — a dynamic it said has historically coincided with further gains in USD/JPY, according to the CNBC report. Changes in Japanese bond supply and U.S. yield levels may remain relevant factors for currency-market participants. 

    The Dollar Side: AI Capex and Energy as Structural Tailwinds

    Goldman’s revised yen call sits inside a broader dollar-bullish framework. The bank attributed its conviction on the greenback to two forces it expects to persist: the U.S. AI investment boom and energy supply disruptions, which it described collectively as a “supply bust” providing structural support for the dollar against lower-yielding currencies, Lee Ying Shan reported for CNBC.

    The bank revised its EUR/USD forecasts lower alongside the yen call, projecting the pair at 1.14 in three months before slipping to 1.12 in six months and holding there over a 12-month horizon, per CNBC. The euro and yen are being painted with the same broad brush — both low-yielding currencies facing a dollar that Goldman expects to stay well supported.

    Pair3-Month Forecast6-Month Forecast12-Month Forecast
    USD/JPY162163165
    EUR/USD1.141.121.12

    Source: Goldman Sachs via CNBC

    Where Goldman Is Still Bullish: High-Carry EM

    The divergence in Goldman’s views is worth registering. While the bank cut its yen and euro outlooks, it strengthened forecasts for the Indian rupee, citing improved growth, lower inflation, and expected capital inflows following Reserve Bank of India measures. It also turned more optimistic on Colombia’s peso following a hawkish central bank stance and expectations of fiscal consolidation, CNBC reported.

    Goldman said it continues to favour using the yen “as a funder for high-carry EM expressions” — the carry trade framework, borrowing in a low-yielding currency to finance positions in higher-yielding markets. This reflects Goldman’s reported view that rate differentials remain an important consideration. 

    What Could Change the Picture

    Goldman is explicit about the two conditions that could break its thesis: a U.S. recession shock, or a materially faster pace of Bank of Japan rate hikes. The bank said intervention may buy time but — absent either of those developments — any support for the yen is likely to prove temporary, per CNBC.

    Both remain plausible tails. A sharper-than-expected U.S. labour market deterioration would reprice the Fed’s path quickly, compressing the yield differential that is doing much of the heavy lifting in Goldman’s dollar call. On the BoJ side, core inflation in Japan has been running above target, and the central bank has shown more willingness to adjust policy than its pre-2024 posture suggested. A shift in BoJ forward guidance — particularly one tied to wage data — could move the pair faster than Goldman’s 12-month path implies.

    Neither scenario is Goldman’s base case. Market participants may continue to monitor Bank of Japan policy statements and U.S. economic data for further context  in the Bank of Japan’s upcoming policy statements and U.S. growth prints on the BLS calendar.

    Risk Disclaimer: Trading CFDs involves substantial risk and may not be suitable for all investors. CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. You may lose some or all of your invested capital. You should consider whether you understand how CFDs work and whether you can afford to take the high risk of losing your money. Past performance is not a reliable indicator of future results. This article is provided for general informational and educational purposes only and does not constitute financial, investment, legal, tax, or trading advice, nor a recommendation, solicitation, or offer to buy or sell any financial instrument.