We are experts in neither Nvidia (NVDA) nor DeepSeek. We are neither endorsing nor seeking to refute the headline above, but given headlines such as this we seek in this post to demonstrate how to use VecViz’s dashboards to explore DeepSeek’s potential impact on the price of NVDA, and then how to layer in the Vector and Sigma models of price probability and V-Score related considerations to arrive at some narrower range of probable impact.
NVDA is currently loaded in VecViz’s dashboards,, with NVDA VecEvents sourced from various popular LLM’s for illustrative purposes, the veracity of which we cannot vouch for. Note that the dashboards are best viewed on laptop/desktop, that they load progressively, and that some of them take 15-30 seconds to fully load.
Brief Overview of Vector Sets, etc..
See Fig1, below, for NVDA”s unfiltered Vector Strength Histogram. The Vector Strength Histogram reflects every channel, or Vector Set, that can be formed from the closing price Tops and Bottoms that VecViz has identified. Each Vector Set includes a “leveled up” and “leveled down” section, of ~ equivalent width. Fibonacci lines, or Vectors, are drawn in each section of each Vector Set. Each such line connects to a bar that together form the Vector Strength Histogram. The length of each bar reflects the “Vector Strength” of the associated vector set. Vector Strength is a function of date and price proximity of the lines comprising it to the current Model Date and price, and the number of “touches” they make with tops and bottoms historically. The dates of the Tops and Bottoms anchoring a Vector Set are its “VecDates”. Key themes and events that are deemed to have influenced the ticker during the time period spanned by a Vector Set’s VecDates are the “VecEvents” for that Vector Set. For more info on these topics see our FAQ and methodology oriented posts in this blog.
Start By Filtering NVDA’s Vector Strength Histogram For DeepSeek Impacted Vector Sets.
The most DeepSeek pertinent NVDA Vector Sets can be identified by filtering the Vector Strength Histogram for the most DeepSeek pertinent VecEvents. If you don’t know NVDA’s VecEvents you can peruse them by hovering your mouse around the Vector Strength Histogram. Start with the biggest bars as they tend to carry the most currently pertinent events. We see the VecEvents linked to “vs1” (Vector Set 1, the “strongest” vector set, as indicated by its bars), detailed below in Fig 2:
Listed among the VecEvents in Fig2, above, for NVDA’s vs1 is the introduction of the H100, which has been reported to be NVDA’s most powerful chip (aside from Blackwell, which is still in early stages of deployment), subject to export controls. For sake of the illustration, let’s proceed as if this premise is correct and focus in on Vector Sets, such as vs1, that include H100 among their VecEvents.
First, as a blunt worst case scenario, we consider in Fig3, below, what the distribution of Vector Strength support and resistance looks like if we exclude all Vector Sets that include the H100 among their VecEvents. To do so we enter “H100” in the text box for VecEvents, put our mouse atop the first listed line of text till it turns grey, and then hold the down arrow on our keyboard until all lines of text are selected, and then we uncheck them. The highest price of the remaining vectors is $94.94, representing the peak levelled up Vector (level 2.236) of vs4 (Vector Set 4), the 4th strongest vector set.
Use the Vector Set Study Area to study DeepSeek impacted Vector Sets
Since vs4 is the strongest Vector Set that does not include the H100 in its VecEvents, presumably vs1, vs2, and vs3 do include the H100, so we plotted them out in the Vector Set Study area in Fig 3, below.
Exploring Fig3 with our mouse in the VecViz Dashboard reveals that NVDA”s price closed on Friday 1/24/25 near the ceiling of the core section (i.e. the 1.00 level) of each of these vector sets, all of which share Jun20204 as an anchoring top. It also reveals that there is some confluence of the center core of these three vector sets (the 0.382 to 0.618 areas) down to around $124.51 that would probably be hard to get past unless bearish NVDA sentiment (DeepSeek related or otherwise) builds further. That said, if DeepSeek pushed NVDA to the core floor of these Vector Sets (i.e. the 0.00 level), where NVDA resided during 2023, NVDA would be in the $102.68/ $84.23 and $78.6 for vs1, vs2, and vs3 respectively.
So, again, it probably takes some real bearish sentiment to get below $124.51 anytime soon, conceptually NVDA could be as low as $102.68 zone and still be trading consistent with the core of its 3 strongest vector sets linked to the H100. At $78.6-$84.23 it would be in the center of the “Levelled Down” section of vs1, signifying a material “shock” likely occurred, but still in the core of of vs2 and vs3. In other words, $102.67 doesn’t suggest a clearly broken H100 growth story, but $84.23-$78.6 might. Between these prices is the aforementioned ceiling of the levelled up section of VS4 at $94.94. Note that VS4’s VecDates end in Nov2021, prior to the introduction of the H100, but still in the CUDA and datacenter era, and that DeepSeek uses NVDA and therefore presumably CUDA and datacenter configurations. NVDA was in the center of this levelled up area of VS4 as recently as April 2024, as depicted in Fig5, below:
Are the outcomes discussed at all probable from a quantitative perspective?
Modestly probable, according to the Vector Model, even less so according to Sigma. As indicated in Fig 6, below, the Support and Resistance / Machine Learning Vector Model puts NVDA at $124.64 as soon as 1d forward from Friday (1/27/25, mislabelled as 1/25/25) with 5% probability. Prices two weeks forward from Friday’s close (2/7/25) at $95.19 and $89.92 with 5% and 1% probability, respectively (see rows “95D” and “99D”). Bell curve parameter-based Sigma, in contrast, doesn’t get to below $125 at the 5% probability level until 2/7/25 ($122.65) and it doesn’t put NVDA at any price below $101.69 with even 1% probability even looking out a month forward (2/22/25).
Does NVDA’s V-Score support the outcomes discussed?
Not as of Friday’s close, for which NVDA’s V-Score was calculated to be a +8 (the full V-Score scale is -12 to +12). That said, +8 is not an exceptionally high V-Score given NVDA’s volatility, and it could change rapidly if its closing price was to register as another “Top” in the near term. For more info on the V-Score see VecViz’s FAQ and this blog entry.
Does NVDA’s V-Score comps support the outcomes discussed?
To some extent, yes, but on balance, as of the close on Friday 1/24/25, no. NVDA’s V-Score comps on 1/24/25 included ISRG, UNH, RIO, OXY, AMZN, NFLX. While ISRG could be viewed as reflecting NVDA’s high tech / AI and robotics focus, and there are arguably parallels between UNH’s Optum and NVDA’s CUDA as analytics ecosystems, OXY and RIO reminding us of the commodity nature of semiconductors more broadly. Note that while OXY fails to outmatch ISRG for the 21d horizon, RIO does outmatch ISRG for the 63d horizon, but as the bullish alternative given the model dates and horizon considered. Moreover, AMZN and NFLX, featuring only as bearish comps based on the model dates and horizons considered for them, reminding us of innovators that came under fairly heavy competitive pressure for periods of time (in the cloud and streaming, respectively). That said, neither was determined to be the better match for NVDA than their bullish counterpart (ISRG prevailed over AMZN and UNH was indeterminate vs NFLX).
To explore these comps further, view the V-Score comparable dashboard for each of the 6 forward time horizons (each yielding an integer V-Score between -2 and +2, or no score at all if the difference in fit cannot be decided by the V-Score machine learning ensemble), an example of which is provided below in Fig7, below.
Conclusion
Our aim here was to use this interesting situation to demonstrate how VecViz’s dashboards can be used to grapple with uncertainty through trajectories, narrative and quantitative techniques. VecViz is not an investment advisor. This was just an illustration. See our Terms and Conditions for important notices and disclosures.
That said, a quick summary of our NVDA / DeepSeek novice reading of the present NVDA situation considered exclusively via VecViz’s dashboards: NVDA carries a fairly bullish outlook heading into this, with a +8 V-Score and fairly solid comps. That said, if NVDA does decline in the week ahead (because of DeepSeek or any other H100 related reason), a decline below $125 carries a 5% probability (calculated as of the Friday, 1/24/25 close) over the next day to couple weeks from the Vector Model and Sigma, respectively, and would be entirely consistent with the H100/premium AI chip vector set growth trajectory . That said, if NVDA does get below $125 in that time frame (merits discussion given that it is in $125 zone pre-market Monday, 1/27/25), downside to the mid $90’s, presently 5% over the next couple weeks according to the Vector Model only, still is not necessarily probable but becomes noticeably large. Mid 90’s in that time frame is not inconsistent with the H100 / premium AI chip vector set core / growth trajectory to any material degree.