DAUGHTREY, J., delivered the opinion in which BATCHELDER, J., joined in the result. BATCHELDER, J. (pp. 741-45), delivered a separate opinion concurring in the judgment. ROGERS, J. (pp. 745-56), delivered a separate dissenting opinion.
OPINION
MARTHA CRAIG DAUGHTREY, Circuit Judge.This case is before us for a second time, following an order of remand in United States v. DTE Energy Co. (DTE I), 711 F.3d 643 (6th Cir. 2013). As we noted there, regulations under the Clean Air Act require a utility seeking to modify a source of air pollutants to “make a preconstruction projection of whether and to what extent emissions from the source will increase following construction.” Id. at 644. This projection then “determines whether the project constitutes a ‘major modification’ and thus requires a permit” prior to construction, as part of the Act’s New Source Review (NSR) program. Id.; see also 42 U.S.C. §§ 7475, 7503; 40 C.F.R. § 52.21. The NSR regulations require an *737operator to “consider all relevant information” when estimating its post-project actual emissions but allow for the exclusion of any emissions “that an existing unit could have accommodated during the [baseline period] ... and that are also unrelated to the particular project, including any increased utilization due to product demand growth.” 40 C.F.R.- § 52.21 (b)(41)(ii)(a) and (c). An operator must document and explain its decision to exclude emissions from its projection as resulting from future “demand growth” ¿nd provide such information to the EPA or to the designated state regulatory agency. 40 C.F.R. § 52.21(r)(6)(i)-(ii).
Defendants DTE Energy Co. and its subsidiary, Detroit Edison Co. (collectively DTE), own and operate the largest coal-fired power plant in Michigan at their facility in Monroe, where, in 2010, DTE undertook a three-month-long overhaul of Unit 2 costing $65 million. On the day before it began construction, DTE submitted a notification to the Michigan Department of Environmental Quality stating that DTE predicted an increase in post-construction emissions 100 times greater than the minimum necessary to constitute a “major modification” and require a pre-construction permit. DTE initially characterized the projects as routine maintenance, repair, and replacement activities, a designation that, if accurate, would exempt the projects from triggering NSR.1 See New York v. U.S. Envtl. Prot. Agency, 443 F.3d 880, 883-84 (D.C. Cir. 2006). DTE also informed the state agency that it had excluded the entire predicted emissions increase from its projections of Unit 2’s post-construction emissions based on “demand growth.” This designation, if it could be established to the agency’s satisfaction, also would have exempted DTE’s modification from the necessity of a permit and, thus, allowed DTE to postpone some of the pollution-control installations that were planned as a future upgrade.2 See 40 C.F.R. § 52.21(b)(41)(ii)(c). DTE began construction on Monroe Unit 2 without obtaining an NSR permit.
After investigation of DTE’s projections, the EPA filed this enforcement action, challenging the company’s routine-maintenance designation and its exclusion for “demand growth,” and insisting that DTE should have secured a preconstruction permit and included pollution controls in the Unit 2 overhaul to remediate the projected emissions increases. The district court granted summary judgment to DTE, holding that the EPA’s enforcement action was premature because the construction had not yet produced an actual increase in emissions. On appeal, we reversed and remanded, holding that the EPA was authorized to bring an enforcement action based on projected increases in emissions without first demonstrating that emissions actually had increased after the project. DTE I, 711 F.3d at 649.
On remand, the district court again entered summary judgment for DTE, this time focusing on language in our first opinion to the effect that “the regulations allow operators to undertake projects without having EPA second-guess their projections.” Id. at 644. The district court appar*738ently (and mistakenly) took this to mean that the EPA had to accept DTE’s projections at face value, holding that:
EPA is only entitled to conduct a surface review of a source operator’s pre-construction projections to determine whether they comport with the letter of the law. Anything beyond this cursory examination would allow EPA to “second-guess” a source operator’s calculations; an avenue which the Sixth Circuit explicitly foreclosed to regulators. [Emphasis added.]
In this case, EPA claims that defendants improperly applied the demand growth exclusion when they “expected pollution from ... Unit 2 to go up by thousands of tons each year after the overhaul,” and then discounted this entire emissions increase by attributing it to additional consumer demand. In other words, EPA does not contend that defendants violated any of the agency’s regulations when they computed the preconstruction emission projections from Unit 2. Rather, EPA takes defendants to task over the extent to which they relied upon the demand growth exclusion to justify their projections. This is exactly what the Sixth Circuit envisioned when it precluded EPA from second-guessing “the making of [precon-struction emission] projections.” [Internal citations omitted.]
The problem with the district court’s analysis is two-fold. First, the focus on so-called “second-guessing” is misplaced. That language from our earlier opinion is, technically speaking, dictum, because the holding of the opinion was, as noted above, that the EPA could bring a preconstruction enforcement action to challenge DTE’s emissions projections. Second, in reviewing an operator’s attribution of increased emissions to demand growth, the EPA definitely is not confined to a “surface review” or “cursory examination.”
Indeed, two agency pronouncements, dating back to 1992, make clear that the EPA must engage in actual review. The first is in 57 Fed. Reg. 32,314, 32,327 (July 21, 1992), which is quoted in our first opinion: “[W]hether the [demand growth] exclusion applies ‘is a fact-dependent determination that must be resolved on a case-by-case basis’ ” DTE I, 711 F.3d at 646 (emphasis added). The second is found in 72 Fed. Reg. 72,607, 72,611 (Dec. 21, 2007) (emphasis added): NSR record-keeping requirements “establish[ ] an adequate paper trail to allow enforcement authorities to evaluate [an operator’s] claims concerning what amount of an emissions increase is related to the project and what amount is attributable to demand growth.”
But the EPA cannot evaluate a fact-dependent claim on a case-by-case basis unless the operator supplies supporting facts, which the record establishes was not done here. In other words, a valid projection must consist of more than the following list, which is, in effect, all that DTE provided to the EPA:
*739Increase in nitrous oxide emissions. 4,096 tons
Increase in sulfur dioxide emissions.. 3,701 tons
Total increase in emissions. .7,797 tons
Less amount attributable to demand growth .7,797 tons
.0 tons NSR projection for post-construction emissions
The record before us is devoid of any support for this thoroughly superficial calculation.3 DTE baldly asserted that it was excluding from its projections “ ‘that portion of the unit’s emissions following the project that an existing unit could have accommodated ... and that are also unrelated to the particular project,’ including increases due to demand and market conditions or fuel quality.” Mar. 12, 2010 Notice Letter, Page ID 165 (quoting the Michigan equivalent of 40 C.F.R. § 52.21(b)(41)(ii)(c)). DTE then went on to claim that “emissions and operations fluctuate year-to-year due to market conditions,” and “[a]t some point in the future, baseline levels may be exceeded again, but not as a result of this outage.” Id. This letter provided no rationale for the company’s claim that Unit 2 was capable of accommodating the increased emissions prior to the construction projects or that future growth in the demand for electricity was the sole cause of the projected increase in pollutants. Although DTE later sent two more letters to the EPA supposedly clarifying the method of calculating baseline emissions, these letters also failed to explain why DTE applied the demand-growth exclusion to its entire projected-emissions increase. In its motion for summary judgment below, DTE claimed that it attributed the increased emissions to future demand for power “[biased on the company’s business and engineering judgment” (Page ID 6716), but gave no specific information to support that judgment.
In fact, not one of DTE’s attempts to justify its application of the demand-growth exclusion was supported by documentation, without which the EPA could not meaningfully evaluate DTE’s projections. There was, in truth, nothing to evaluate. Moreover, the results of a computer model that DTE ran, when it was rerun by the EPA, showed that DTE should actually have predicted a decrease in demand. (Page ID 372) Contrary to DTE’s “business and engineering judgment,” what did occur in the immediate post-construction period was a decline in consumer demand, not an increase. Appellee’s Br. at 64.
DTE’s failure to carry its burden to set out a factual basis for its demand-growth exclusion is just one problem with its projections. In order to exclude increased emissions as the product of increased demand under 40 C.F.R. § 52.21(b)(41)(ii), the company must establish (1) that the projected post-construction emissions could have been accommodated during the preconstruction period and (2) that the projected emissions are unrelated to the construction project.4 As to the first re*740quirement, DTE did not and could not establish that the increase in emissions could have been accommodated during the baseline period. Prior to the overhaul, DTE was running Unit 2 at full capacity— that is, Unit 2 was operating every hour that it could be operated. (Page ID 294) But Unit 2 was experiencing continual outages that kept it from running almost 20 percent of the time (Page ID 302), which is obviously why DTE shut it down for three months to accomplish the overhaul, aimed at increasing efficiency and reliability. For the same reason, DTE did not and could not establish that the increase in emissions was unrelated to the construction process. The planned increase in efficiency and reliability would allow the plant to operate for at least an additional 12 days each year (Page ID 306), which in turn would result in increased emissions unless the construction also had included pollution controls, as the issuance of a permit would have required.
In DTE I, we referenced the second sentence of 40 C.F.R. § 52.21(r)(6)(ii):
If the emissions unit is an existing electric utility steam generating unit, before beginning actual construction, the owner or operator shall provide a copy of the information set out in paragraph (r)(6)(i). Nothing in this paragraph (r)(6)(ii) shall be construed to require the owner or operator of such a unit to obtain any determination from the Administrator before beginning actual construction.
711 F.3d at 650 (emphasis added). Judge Rogers’s current dissent seems to take a broader view of this regulation than the text permits in repeatedly cautioning that permitting the EPA’s enforcement action to go forward would create “a de facto prior approval system.” (Rogers Opinion at 745-46, 747, 748-49) But this reading is patently too expansive, because the regulation does not say that the EPA has to accept projections at face value or that it is prohibited from questioning their legitimacy. Instead, and in context, the rule means that once the required information has been submitted to the EPA for review, the operator does not have to delay construction until it receives a decision on the necessity of a permit, but may commence construction prior to a “determination from the Administrator.” Of course, if the operator actually begins construction without waiting for a “determination” from the EPA and it later turns out that a permit was required, a violation of NSR has occurred, and the operator risks penalties and injunctive relief requiring mitigation of illegal emissions, a possible shut down of the unit, or a retrofit with pollution controls to meet emissions standards. See, e.g., United States v. Cinergy Corp., 618 F.Supp.2d 942, 971 (S.D. Ind. 2009), rev’d on other grounds, 623 F.3d 455 (7th Cir. 2010).
In short, DTE was not required by the regulations to secure the EPA’s approval of the projections, or the project, before beginning construction, but in going forward without a permit, DTE proceeded at its own risk. The EPA is not prevented by law or by our prior opinion in DTE I from challenging DTE’s preconstruction projections, such as they are. Viewing the facts in the light most favorable to the EPA, we conclude that there are genuine disputes of *741material fact that preclude summary judgment for DTE regarding DTE’s compliance with NSR’s statutory preconstruction requirements and with agency regulations implementing those provisions. Therefore, we REVERSE the district court’s grant of summary judgment to DTE and REMAND this case for further proceedings consistent with this opinion.
In terms of the remand, it is important to note that the panel unanimously agrees — now that DTE I is the law of this case and of the circuit — that actual post-construction emissions have no bearing on the question of whether DTE’s precon-struction projections complied with the regulations. (Batchelder Concurrence at 745; Rogers Opinion at 749) DTE I foreclosed that question in holding that an operator who begins construction without making a projection in accordance with the regulations is subject to enforcement, no matter what post-construction data later shows. 711 F.3d at 649. The district court erred initially and again on remand when it ruled that post-construction data could be used to show that a construction project was not a “major modification.” Apparently, it is necessary to reiterate that the applicability of NSR must be determined before construction commences and that liability can attach if an operator proceeds to construction without complying with the preconstruction requirements in the regulations. Post-construction emissions data cannot prevent the EPA from challenging DTE’s failure to comply with NSR’s pre-construction requirements.
. As it turns out, the EPA does not consider a $65-million overhaul to be routine by definition, .
. Those upgrades have since been completed. Since the Monroe Unit 2 overhaul was completed in 2010, DTE has installed the scrubbers and other pollution controls necessary to remediate toxic emissions at the facility, so that implementation is no longer at issue. Appellee's Br. at 13 n.4. But, if it is found to have violated the Act, DTE still could face monetary penalties and be required to mitigate excess emissions caused by the delay in installing pollution controls.
. Clearly, DTE failed to comply with the regulation requiring it to “document ... the amount of emissions excluded under paragraph (b)(41)(ii)(c) of this section and an explanation for why such amount was excluded.” 40 C.F.R. § 52.21 (r)(6)(i).
. Both requirements must be met. See New York v. U.S. Envtl. Prot. Agency, 413 F.3d 3, *74033 (D.C. Cir. 2005) (citing 67 Fed. Reg. 80,186, 80,203 (Dec. 31, 2002)) ("[E]ven if the operation of an emissions unit to meet a particular level of demand could have been accomplished during the representative baseline period, but it can be shown that the increase is related to the changes made to the unit, then the emissions increases resulting from the increased operation must be attributed to the modification project, and cannot be subtracted from the projection of post-change actual emissions.”).